Datasets:

NAACL

/

NAACL2024-papers

like 0

Follow

North American Chapter of the Association for Computational Linguistics 1

https://aclanthology.org/2024.trustnlp-1.7.bib

https://aclanthology.org/2024.trustnlp-1.7/

@inproceedings{yoo-etal-2024-exploring, title = "Exploring Causal Mechanisms for Machine Text Detection Methods", author = "Yoo, Kiyoon and Ahn, Wonhyuk and Song, Yeji and Kwak, Nojun", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.7", doi = "10.18653/v1/2024.trustnlp-1.7", pages = "71--78", abstract = "The immense attraction towards text generation garnered by ChatGPT has spurred the need for discriminating machine-text from human text. In this work, we provide preliminary evidence that the scores computed by existing zero-shot and supervised machine-generated text detection methods are not solely determined by the generated texts, but are affected by prompts and real texts as well. Using techniques from causal inference, we show the existence of backdoor paths that confounds the relationships between text and its detection score and how the confounding bias can be partially mitigated. We open up new research directions in identifying other factors that may be interwoven in the detection of machine text. Our study calls for a deeper investigation into which kinds of prompts make the detection of machine text more difficult or easier", }

The immense attraction towards text generation garnered by ChatGPT has spurred the need for discriminating machine-text from human text. In this work, we provide preliminary evidence that the scores computed by existing zero-shot and supervised machine-generated text detection methods are not solely determined by the generated texts, but are affected by prompts and real texts as well. Using techniques from causal inference, we show the existence of backdoor paths that confounds the relationships between text and its detection score and how the confounding bias can be partially mitigated. We open up new research directions in identifying other factors that may be interwoven in the detection of machine text. Our study calls for a deeper investigation into which kinds of prompts make the detection of machine text more difficult or easier

Exploring Causal Mechanisms for Machine Text Detection Methods

trustnlp-1.7

Poster

https://aclanthology.org/2024.trustnlp-1.8.bib

https://aclanthology.org/2024.trustnlp-1.8/

@inproceedings{rashad-etal-2024-factalign, title = "{F}act{A}lign: Fact-Level Hallucination Detection and Classification Through Knowledge Graph Alignment", author = "Rashad, Mohamed and Zahran, Ahmed and Amin, Abanoub and Abdelaal, Amr and Altantawy, Mohamed", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.8", doi = "10.18653/v1/2024.trustnlp-1.8", pages = "79--84", abstract = "This paper proposes a novel black-box approach for fact-level hallucination detection and classification by transforming the problem into a knowledge graph alignment task. This approach allows us to classify detected hallucinations as either intrinsic or extrinsic. The paper starts by discussing the field of hallucination detection and introducing several approaches to related work. Then, we introduce the proposed FactAlign approach for hallucination detection and discuss how we can use it to classify hallucinations as either intrinsic or extrinsic. Experiments are carried out to evaluate the proposed method against state-of-the-art methods on the hallucination detection task using the WikiBio GPT-3 hallucination dataset, and on the hallucination type classification task using the XSum hallucination annotations dataset. The experimental results show that our method achieves a 0.889 F1 score for the hallucination detection and 0.825 F1 for the hallucination type classification, without any further training, fine-tuning, or producing multiple samples of the LLM response.", }

This paper proposes a novel black-box approach for fact-level hallucination detection and classification by transforming the problem into a knowledge graph alignment task. This approach allows us to classify detected hallucinations as either intrinsic or extrinsic. The paper starts by discussing the field of hallucination detection and introducing several approaches to related work. Then, we introduce the proposed FactAlign approach for hallucination detection and discuss how we can use it to classify hallucinations as either intrinsic or extrinsic. Experiments are carried out to evaluate the proposed method against state-of-the-art methods on the hallucination detection task using the WikiBio GPT-3 hallucination dataset, and on the hallucination type classification task using the XSum hallucination annotations dataset. The experimental results show that our method achieves a 0.889 F1 score for the hallucination detection and 0.825 F1 for the hallucination type classification, without any further training, fine-tuning, or producing multiple samples of the LLM response.

FactAlign: Fact-Level Hallucination Detection and Classification Through Knowledge Graph Alignment

trustnlp-1.8

Poster

https://aclanthology.org/2024.trustnlp-1.9.bib

https://aclanthology.org/2024.trustnlp-1.9/

@inproceedings{fu-etal-2024-cross, title = "Cross-Task Defense: Instruction-Tuning {LLM}s for Content Safety", author = "Fu, Yu and Xiao, Wen and Chen, Jia and Li, Jiachen and Papalexakis, Evangelos and Chien, Aichi and Dong, Yue", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.9", doi = "10.18653/v1/2024.trustnlp-1.9", pages = "85--93", abstract = "Recent studies reveal that Large Language Models (LLMs) face challenges in balancing safety with utility, particularly when processing long texts for NLP tasks like summarization and translation. Despite defenses against malicious short questions, the ability of LLMs to safely handle dangerous long content, such as manuals teaching illicit activities, remains unclear. Our work aims to develop robust defenses for LLMs in processing malicious documents alongside benign NLP task queries. We introduce a defense dataset comprised of safety-related examples and propose single-task and mixed-task losses for instruction tuning. Our empirical results demonstrate that LLMs can significantly enhance their capacity to safely manage dangerous content with appropriate instruction tuning. Additionally, strengthening the defenses of tasks most susceptible to misuse is effective in protecting LLMs against processing harmful information. We also observe that trade-offs between utility and safety exist in defense strategies, where Llama2, utilizing our proposed approach, displays a significantly better balance compared to Llama1.", }

Recent studies reveal that Large Language Models (LLMs) face challenges in balancing safety with utility, particularly when processing long texts for NLP tasks like summarization and translation. Despite defenses against malicious short questions, the ability of LLMs to safely handle dangerous long content, such as manuals teaching illicit activities, remains unclear. Our work aims to develop robust defenses for LLMs in processing malicious documents alongside benign NLP task queries. We introduce a defense dataset comprised of safety-related examples and propose single-task and mixed-task losses for instruction tuning. Our empirical results demonstrate that LLMs can significantly enhance their capacity to safely manage dangerous content with appropriate instruction tuning. Additionally, strengthening the defenses of tasks most susceptible to misuse is effective in protecting LLMs against processing harmful information. We also observe that trade-offs between utility and safety exist in defense strategies, where Llama2, utilizing our proposed approach, displays a significantly better balance compared to Llama1.

Cross-Task Defense: Instruction-Tuning LLMs for Content Safety

trustnlp-1.9

Poster

2405.15202

https://aclanthology.org/2024.trustnlp-1.10.bib

https://aclanthology.org/2024.trustnlp-1.10/

@inproceedings{brandl-etal-2024-interplay, title = "On the Interplay between Fairness and Explainability", author = "Brandl, Stephanie and Bugliarello, Emanuele and Chalkidis, Ilias", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.10", doi = "10.18653/v1/2024.trustnlp-1.10", pages = "94--108", abstract = "In order to build reliable and trustworthy NLP applications, models need to be both fair across different demographics and explainable. Usually these two objectives, fairness and explainability, are optimized and/or examined independently of each other. Instead, we argue that forthcoming, trustworthy NLP systems should consider both.In this work, we perform a first study to understand how they influence each other: do fair(er) models rely on more plausible explanations? and vice versa. To this end, we conduct experiments on two English multi-class text classification datasets, BIOS and ECtHR, that provide information on gender and nationality, respectively, as well as human-annotated rationales. We fine-tune pre-trained language models with several methods for (i) bias mitigation, which aims to improve fairness; (ii) rationale extraction, which aims to produce plausible explanations.We find that bias mitigation algorithms do not always lead to fairer models. Moreover, in our analysis, we see that empirical fairness and explainability are orthogonal.", }

In order to build reliable and trustworthy NLP applications, models need to be both fair across different demographics and explainable. Usually these two objectives, fairness and explainability, are optimized and/or examined independently of each other. Instead, we argue that forthcoming, trustworthy NLP systems should consider both.In this work, we perform a first study to understand how they influence each other: do fair(er) models rely on more plausible explanations? and vice versa. To this end, we conduct experiments on two English multi-class text classification datasets, BIOS and ECtHR, that provide information on gender and nationality, respectively, as well as human-annotated rationales. We fine-tune pre-trained language models with several methods for (i) bias mitigation, which aims to improve fairness; (ii) rationale extraction, which aims to produce plausible explanations.We find that bias mitigation algorithms do not always lead to fairer models. Moreover, in our analysis, we see that empirical fairness and explainability are orthogonal.

On the Interplay between Fairness and Explainability

trustnlp-1.10

Poster

2310.16607

https://aclanthology.org/2024.trustnlp-1.11.bib

https://aclanthology.org/2024.trustnlp-1.11/

@inproceedings{cecchini-etal-2024-holistic, title = "Holistic Evaluation of Large Language Models: Assessing Robustness, Accuracy, and Toxicity for Real-World Applications", author = "Cecchini, David and Nazir, Arshaan and Chakravarthy, Kalyan and Kocaman, Veysel", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.11", doi = "10.18653/v1/2024.trustnlp-1.11", pages = "109--117", abstract = "Large Language Models (LLMs) have been widely used in real-world applications. However, as LLMs evolve and new datasets are released, it becomes crucial to build processes to evaluate and control the models{'} performance. In this paper, we describe how to add Robustness, Accuracy, and Toxicity scores to model comparison tables, or leaderboards. We discuss the evaluation metrics, the approaches considered, and present the results of the first evaluation round for model Robustness, Accuracy, and Toxicity scores. Our results show that GPT 4 achieves top performance on robustness and accuracy test, while Llama 2 achieves top performance on the toxicity test. We note that newer open-source models such as open chat 3.5 and neural chat 7B can perform well on these three test categories. Finally, domain-specific tests and models are also planned to be added to the leaderboard to allow for a more detailed evaluation of models in specific areas such as healthcare, legal, and finance.", }

Large Language Models (LLMs) have been widely used in real-world applications. However, as LLMs evolve and new datasets are released, it becomes crucial to build processes to evaluate and control the models{'} performance. In this paper, we describe how to add Robustness, Accuracy, and Toxicity scores to model comparison tables, or leaderboards. We discuss the evaluation metrics, the approaches considered, and present the results of the first evaluation round for model Robustness, Accuracy, and Toxicity scores. Our results show that GPT 4 achieves top performance on robustness and accuracy test, while Llama 2 achieves top performance on the toxicity test. We note that newer open-source models such as open chat 3.5 and neural chat 7B can perform well on these three test categories. Finally, domain-specific tests and models are also planned to be added to the leaderboard to allow for a more detailed evaluation of models in specific areas such as healthcare, legal, and finance.

Holistic Evaluation of Large Language Models: Assessing Robustness, Accuracy, and Toxicity for Real-World Applications

trustnlp-1.11

Poster

https://aclanthology.org/2024.trustnlp-1.12.bib

https://aclanthology.org/2024.trustnlp-1.12/

@inproceedings{fang-etal-2024-hgot, title = "{HGOT}: Hierarchical Graph of Thoughts for Retrieval-Augmented In-Context Learning in Factuality Evaluation", author = "Fang, Yihao and Thomas, Stephen and Zhu, Xiaodan", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.12", doi = "10.18653/v1/2024.trustnlp-1.12", pages = "118--144", abstract = "With the widespread adoption of large language models (LLMs) in numerous applications, the challenge of factuality and the propensity for hallucinations has emerged as a significant concern. To address this issue, particularly in retrieval-augmented in-context learning, we introduce the hierarchical graph of thoughts (HGOT), a structured, multi-layered graph approach designed to enhance the retrieval of pertinent passages during in-context learning. The framework utilizes the emergent planning capabilities of LLMs, employing the divide-and-conquer strategy to break down complex queries into manageable sub-queries. It refines self-consistency majority voting for answer selection, which incorporates the recently proposed citation recall and precision metrics to assess the quality of thoughts, linking an answer{'}s credibility intrinsically to the thought{'}s quality. This methodology introduces a weighted system in majority voting, prioritizing answers based on the citation quality of their thoughts. Additionally, we propose a scoring mechanism for evaluating retrieved passages, considering factors such as citation frequency and quality, self-consistency confidence, and the retrieval module{'}s ranking. Experiments indicate that HGOT excels as a versatile approach, outperforming competing models in FEVER by up to 7{\%} and matching leading models such as Retrieve-then-Read in Open-SQuAD, and DSP in HotPotQA, demonstrating its efficacy in enhancing LLMs{'} factuality.", }

With the widespread adoption of large language models (LLMs) in numerous applications, the challenge of factuality and the propensity for hallucinations has emerged as a significant concern. To address this issue, particularly in retrieval-augmented in-context learning, we introduce the hierarchical graph of thoughts (HGOT), a structured, multi-layered graph approach designed to enhance the retrieval of pertinent passages during in-context learning. The framework utilizes the emergent planning capabilities of LLMs, employing the divide-and-conquer strategy to break down complex queries into manageable sub-queries. It refines self-consistency majority voting for answer selection, which incorporates the recently proposed citation recall and precision metrics to assess the quality of thoughts, linking an answer{'}s credibility intrinsically to the thought{'}s quality. This methodology introduces a weighted system in majority voting, prioritizing answers based on the citation quality of their thoughts. Additionally, we propose a scoring mechanism for evaluating retrieved passages, considering factors such as citation frequency and quality, self-consistency confidence, and the retrieval module{'}s ranking. Experiments indicate that HGOT excels as a versatile approach, outperforming competing models in FEVER by up to 7{\%} and matching leading models such as Retrieve-then-Read in Open-SQuAD, and DSP in HotPotQA, demonstrating its efficacy in enhancing LLMs{'} factuality.

HGOT: Hierarchical Graph of Thoughts for Retrieval-Augmented In-Context Learning in Factuality Evaluation

trustnlp-1.12

Poster

2402.09390

https://aclanthology.org/2024.trustnlp-1.13.bib

https://aclanthology.org/2024.trustnlp-1.13/

@inproceedings{groot-valdenegro-toro-2024-overconfidence, title = "Overconfidence is Key: Verbalized Uncertainty Evaluation in Large Language and Vision-Language Models", author = "Groot, Tobias and Valdenegro - Toro, Matias", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.13", doi = "10.18653/v1/2024.trustnlp-1.13", pages = "145--171", abstract = "Language and Vision-Language Models (LLMs/VLMs) have revolutionized the field of AI by their ability to generate human-like text and understand images, but ensuring their reliability is crucial. This paper aims to evaluate the ability of LLMs (GPT4, GPT-3.5, LLaMA2, and PaLM 2) and VLMs (GPT4V and Gemini Pro Vision) to estimate their verbalized uncertainty via prompting. We propose the new Japanese Uncertain Scenes (JUS) dataset, aimed at testing VLM capabilities via difficult queries and object counting, and the Net Calibration Error (NCE) to measure direction of miscalibration.Results show that both LLMs and VLMs have a high calibration error and are overconfident most of the time, indicating a poor capability for uncertainty estimation. Additionally we develop prompts for regression tasks, and we show that VLMs have poor calibration when producing mean/standard deviation and 95{\%} confidence intervals.", }

Language and Vision-Language Models (LLMs/VLMs) have revolutionized the field of AI by their ability to generate human-like text and understand images, but ensuring their reliability is crucial. This paper aims to evaluate the ability of LLMs (GPT4, GPT-3.5, LLaMA2, and PaLM 2) and VLMs (GPT4V and Gemini Pro Vision) to estimate their verbalized uncertainty via prompting. We propose the new Japanese Uncertain Scenes (JUS) dataset, aimed at testing VLM capabilities via difficult queries and object counting, and the Net Calibration Error (NCE) to measure direction of miscalibration.Results show that both LLMs and VLMs have a high calibration error and are overconfident most of the time, indicating a poor capability for uncertainty estimation. Additionally we develop prompts for regression tasks, and we show that VLMs have poor calibration when producing mean/standard deviation and 95{\%} confidence intervals.

Overconfidence is Key: Verbalized Uncertainty Evaluation in Large Language and Vision-Language Models

trustnlp-1.13

Poster

2405.02917

https://aclanthology.org/2024.trustnlp-1.14.bib

https://aclanthology.org/2024.trustnlp-1.14/

@inproceedings{patel-etal-2024-tweak, title = "Tweak to Trust: Assessing the Reliability of Summarization Metrics in Contact Centers via Perturbed Summaries", author = "Patel, Kevin and Agrawal, Suraj and Kumar, Ayush", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.14", doi = "10.18653/v1/2024.trustnlp-1.14", pages = "172--186", abstract = "In the dynamic realm of call center communications, the potential of abstractive summarization to transform information condensation is evident. However, evaluating the performance of abstractive summarization systems within contact center domain poses a significant challenge. Traditional evaluation metrics prove inadequate in capturing the multifaceted nature of call center conversations, characterized by diverse topics, emotional nuances, and dynamic contexts. This paper uses domain-specific perturbed summaries to scrutinize the robustness of summarization metrics in the call center domain. Through extensive experiments on call center data, we illustrate how perturbed summaries uncover limitations in existing metrics. We additionally utilize perturbation as data augmentation strategy to train domain-specific metrics. Our findings underscore the potential of perturbed summaries to complement current evaluation techniques, advancing reliable and adaptable summarization solutions in the call center domain.", }

In the dynamic realm of call center communications, the potential of abstractive summarization to transform information condensation is evident. However, evaluating the performance of abstractive summarization systems within contact center domain poses a significant challenge. Traditional evaluation metrics prove inadequate in capturing the multifaceted nature of call center conversations, characterized by diverse topics, emotional nuances, and dynamic contexts. This paper uses domain-specific perturbed summaries to scrutinize the robustness of summarization metrics in the call center domain. Through extensive experiments on call center data, we illustrate how perturbed summaries uncover limitations in existing metrics. We additionally utilize perturbation as data augmentation strategy to train domain-specific metrics. Our findings underscore the potential of perturbed summaries to complement current evaluation techniques, advancing reliable and adaptable summarization solutions in the call center domain.

Tweak to Trust: Assessing the Reliability of Summarization Metrics in Contact Centers via Perturbed Summaries

trustnlp-1.14

Poster

https://aclanthology.org/2024.trustnlp-1.15.bib

https://aclanthology.org/2024.trustnlp-1.15/

@inproceedings{khalatbari-etal-2024-flatness, title = "Flatness-Aware Gradient Descent for Safe Conversational {AI}", author = "Khalatbari, Leila and Hosseini, Saeid and Sameti, Hossein and Fung, Pascale", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.15", doi = "10.18653/v1/2024.trustnlp-1.15", pages = "187--195", abstract = "As generative dialog models become ubiquitous in real-world applications, it is paramount to ensure a harmless generation. There are two major challenges when enforcing safety to open-domain chatbots. Firstly, it is impractical to provide training data reflecting the desired response to all emerging forms of toxicity (generalisation challenge). Secondly, implementing safety features may compromise the quality of the conversation (trade-off challenge). To tackle the challenges, this paper introduces a regularized fine-tuning approach called FlatGD. By employing a safety-tailored loss, we translate better optimization to more safety. To ensure better optimization, FlatGD penalizes sharp trajectories of loss curve, encouraging flatness of the converged local minima. Experimental results on datasets of {``}BAD{''} and {``}prosocial dialog{''} demonstrate that our model outperforms the current baselines in reducing toxicity while preserving the conversation quality. Moreover, compared to other baselines, FlatGD can better generalize to unseen toxic data.", }

As generative dialog models become ubiquitous in real-world applications, it is paramount to ensure a harmless generation. There are two major challenges when enforcing safety to open-domain chatbots. Firstly, it is impractical to provide training data reflecting the desired response to all emerging forms of toxicity (generalisation challenge). Secondly, implementing safety features may compromise the quality of the conversation (trade-off challenge). To tackle the challenges, this paper introduces a regularized fine-tuning approach called FlatGD. By employing a safety-tailored loss, we translate better optimization to more safety. To ensure better optimization, FlatGD penalizes sharp trajectories of loss curve, encouraging flatness of the converged local minima. Experimental results on datasets of {``}BAD{''} and {``}prosocial dialog{''} demonstrate that our model outperforms the current baselines in reducing toxicity while preserving the conversation quality. Moreover, compared to other baselines, FlatGD can better generalize to unseen toxic data.

Flatness-Aware Gradient Descent for Safe Conversational AI

trustnlp-1.15

Poster

https://aclanthology.org/2024.trustnlp-1.16.bib

https://aclanthology.org/2024.trustnlp-1.16/

@inproceedings{cao-etal-2024-introducing, title = "Introducing {G}en{C}eption for Multimodal {LLM} Benchmarking: You May Bypass Annotations", author = "Cao, Lele and Buchner, Valentin and Senane, Zineb and Yang, Fangkai", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.16", doi = "10.18653/v1/2024.trustnlp-1.16", pages = "196--201", abstract = "Multimodal Large Language Models (MLLMs) are commonly evaluated using costly annotated multimodal benchmarks. However, these benchmarks often struggle to keep pace with the rapidly advancing requirements of MLLM evaluation. We propose GenCeption, a novel and annotation-free MLLM evaluation framework that merely requires unimodal data to assess inter-modality semantic coherence and inversely reflects the models{'} inclination to hallucinate. Analogous to the popular DrawCeption game, GenCeption initiates with a non-textual sample and undergoes a series of iterative description and generation steps. Semantic drift across iterations is quantified using the GC@T metric. Our empirical findings validate GenCeption{'}s efficacy, showing strong correlations with popular MLLM benchmarking results. GenCeption may be extended to mitigate training data contamination by utilizing ubiquitous, previously unseen unimodal data.", }

Multimodal Large Language Models (MLLMs) are commonly evaluated using costly annotated multimodal benchmarks. However, these benchmarks often struggle to keep pace with the rapidly advancing requirements of MLLM evaluation. We propose GenCeption, a novel and annotation-free MLLM evaluation framework that merely requires unimodal data to assess inter-modality semantic coherence and inversely reflects the models{'} inclination to hallucinate. Analogous to the popular DrawCeption game, GenCeption initiates with a non-textual sample and undergoes a series of iterative description and generation steps. Semantic drift across iterations is quantified using the GC@T metric. Our empirical findings validate GenCeption{'}s efficacy, showing strong correlations with popular MLLM benchmarking results. GenCeption may be extended to mitigate training data contamination by utilizing ubiquitous, previously unseen unimodal data.

Introducing GenCeption for Multimodal LLM Benchmarking: You May Bypass Annotations

trustnlp-1.16

Poster

https://aclanthology.org/2024.trustnlp-1.17.bib

https://aclanthology.org/2024.trustnlp-1.17/

@inproceedings{gao-etal-2024-semantic, title = "Semantic-Preserving Adversarial Example Attack against {BERT}", author = "Gao, Chongyang and Gu, Kang and Vosoughi, Soroush and Mehnaz, Shagufta", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.17", doi = "10.18653/v1/2024.trustnlp-1.17", pages = "202--207", abstract = "Adversarial example attacks against textual data have been drawing increasing attention in both the natural language processing (NLP) and security domains. However, most of the existing attacks overlook the importance of semantic similarity and yield easily recognizable adversarial samples. As a result, the defense methods developed in response to these attacks remain vulnerable and could be evaded by advanced adversarial examples that maintain high semantic similarity with the original, non-adversarial text. Hence, this paper aims to investigate the extent of textual adversarial examples in maintaining such high semantic similarity. We propose Reinforce attack, a reinforcement learning-based framework to generate adversarial text that preserves high semantic similarity with the original text. In particular, the attack process is controlled by a reward function rather than heuristics, as in previous methods, to encourage higher semantic similarity and lower query costs. Through automatic and human evaluations, we show that our generated adversarial texts preserve significantly higher semantic similarity than state-of-the-art attacks while achieving similar attack success rates (outperforming at times), thus uncovering novel challenges for effective defenses.", }

Adversarial example attacks against textual data have been drawing increasing attention in both the natural language processing (NLP) and security domains. However, most of the existing attacks overlook the importance of semantic similarity and yield easily recognizable adversarial samples. As a result, the defense methods developed in response to these attacks remain vulnerable and could be evaded by advanced adversarial examples that maintain high semantic similarity with the original, non-adversarial text. Hence, this paper aims to investigate the extent of textual adversarial examples in maintaining such high semantic similarity. We propose Reinforce attack, a reinforcement learning-based framework to generate adversarial text that preserves high semantic similarity with the original text. In particular, the attack process is controlled by a reward function rather than heuristics, as in previous methods, to encourage higher semantic similarity and lower query costs. Through automatic and human evaluations, we show that our generated adversarial texts preserve significantly higher semantic similarity than state-of-the-art attacks while achieving similar attack success rates (outperforming at times), thus uncovering novel challenges for effective defenses.

Semantic-Preserving Adversarial Example Attack against BERT

trustnlp-1.17

Poster

https://aclanthology.org/2024.trustnlp-1.18.bib

https://aclanthology.org/2024.trustnlp-1.18/

@inproceedings{upadhayay-behzadan-2024-sandwich, title = "Sandwich attack: Multi-language Mixture Adaptive Attack on {LLM}s", author = "Upadhayay, Bibek and Behzadan, Vahid", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.18", doi = "10.18653/v1/2024.trustnlp-1.18", pages = "208--226", abstract = "A significant challenge in reliable deployment of Large Language Models (LLMs) is malicious manipulation via adversarial prompting techniques such as jailbreaks. Employing mechanisms such as safety training have proven useful in addressing this challenge. However, in multilingual LLMs, adversaries can exploit the imbalanced representation of low-resource languages in datasets used for pretraining and safety training. In this paper, we introduce a new black-box attack vector called the Sandwich Attack: a multi-language mixture attack, which manipulates state-of-the-art LLMs into generating harmful and misaligned responses. Our experiments with five different models, namely Bard, Gemini Pro, LLaMA-2-70-B-Chat, GPT-3.5-Turbo, GPT-4, and Claude-3-OPUS, show that this attack vector can be used by adversaries to elicit harmful responses from these models. By detailing both the mechanism and impact of the Sandwich attack, this paper aims to guide future research and development towards more secure and resilient LLMs, ensuring they serve the public good while minimizing potential for misuse. Content Warning: This paper contains examples of harmful language.", }

A significant challenge in reliable deployment of Large Language Models (LLMs) is malicious manipulation via adversarial prompting techniques such as jailbreaks. Employing mechanisms such as safety training have proven useful in addressing this challenge. However, in multilingual LLMs, adversaries can exploit the imbalanced representation of low-resource languages in datasets used for pretraining and safety training. In this paper, we introduce a new black-box attack vector called the Sandwich Attack: a multi-language mixture attack, which manipulates state-of-the-art LLMs into generating harmful and misaligned responses. Our experiments with five different models, namely Bard, Gemini Pro, LLaMA-2-70-B-Chat, GPT-3.5-Turbo, GPT-4, and Claude-3-OPUS, show that this attack vector can be used by adversaries to elicit harmful responses from these models. By detailing both the mechanism and impact of the Sandwich attack, this paper aims to guide future research and development towards more secure and resilient LLMs, ensuring they serve the public good while minimizing potential for misuse. Content Warning: This paper contains examples of harmful language.

Sandwich attack: Multi-language Mixture Adaptive Attack on LLMs

trustnlp-1.18

Poster

2404.07242

https://aclanthology.org/2024.trustnlp-1.19.bib

https://aclanthology.org/2024.trustnlp-1.19/

@inproceedings{yang-etal-2024-masking, title = "Masking Latent Gender Knowledge for Debiasing Image Captioning", author = "Yang, Fan and Ghosh, Shalini and Barut, Emre and Qin, Kechen and Wanigasekara, Prashan and Su, Chengwei and Ruan, Weitong and Gupta, Rahul", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.19", doi = "10.18653/v1/2024.trustnlp-1.19", pages = "227--238", abstract = "Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99{\%} gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics.", }

Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99{\%} gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics.

Masking Latent Gender Knowledge for Debiasing Image Captioning

trustnlp-1.19

Poster

https://aclanthology.org/2024.trustnlp-1.20.bib

https://aclanthology.org/2024.trustnlp-1.20/

@inproceedings{bauer-etal-2024-believe, title = "{BELIEVE}: Belief-Enhanced Instruction Generation and Augmentation for Zero-Shot Bias Mitigation", author = "Bauer, Lisa and Mehrabi, Ninareh and Goyal, Palash and Chang, Kai-Wei and Galstyan, Aram and Gupta, Rahul", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.20", doi = "10.18653/v1/2024.trustnlp-1.20", pages = "239--251", abstract = "Language models, pre-trained on large amounts of unmoderated content, have been shown to contain societal biases. Mitigating such biases typically requires access to model parameters and training schemas. In this work, we address bias mitigation at inference time, such that it can be applied to any black-box model. To this end, we propose a belief generation and augmentation framework, BELIEVE, that demonstrates effective bias mitigation for natural language generation by augmenting input prompts with automatically generated instruction-based beliefs. Our framework eases the bottleneck required for manually crafting these instruction-based beliefs, by extending a recently proposed iterative in-context learning framework to automatically generate beliefs via a language model. We assess the impact of this system on fairness, and demonstrate effective bias mitigation on pretrained and instruction-tuned models for both sentiment and regard with respect to multiple protected classes including race, gender, and political ideology.", }

Language models, pre-trained on large amounts of unmoderated content, have been shown to contain societal biases. Mitigating such biases typically requires access to model parameters and training schemas. In this work, we address bias mitigation at inference time, such that it can be applied to any black-box model. To this end, we propose a belief generation and augmentation framework, BELIEVE, that demonstrates effective bias mitigation for natural language generation by augmenting input prompts with automatically generated instruction-based beliefs. Our framework eases the bottleneck required for manually crafting these instruction-based beliefs, by extending a recently proposed iterative in-context learning framework to automatically generate beliefs via a language model. We assess the impact of this system on fairness, and demonstrate effective bias mitigation on pretrained and instruction-tuned models for both sentiment and regard with respect to multiple protected classes including race, gender, and political ideology.

BELIEVE: Belief-Enhanced Instruction Generation and Augmentation for Zero-Shot Bias Mitigation

trustnlp-1.20

Poster

https://aclanthology.org/2024.trustnlp-1.21.bib

https://aclanthology.org/2024.trustnlp-1.21/

@inproceedings{zarharan-etal-2024-tell, title = "Tell Me Why: Explainable Public Health Fact-Checking with Large Language Models", author = "Zarharan, Majid and Wullschleger, Pascal and Behkam Kia, Babak and Pilehvar, Mohammad Taher and Foster, Jennifer", editor = "Ovalle, Anaelia and Chang, Kai-Wei and Cao, Yang Trista and Mehrabi, Ninareh and Zhao, Jieyu and Galstyan, Aram and Dhamala, Jwala and Kumar, Anoop and Gupta, Rahul", booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.trustnlp-1.21", doi = "10.18653/v1/2024.trustnlp-1.21", pages = "252--278", abstract = "This paper presents a comprehensive analysis of explainable fact-checking through a series of experiments, focusing on the ability of large language models to verify public health claims and provide explanations or justifications for their veracity assessments. We examine the effectiveness of zero/few-shot prompting and parameter-efficient fine-tuning across various open and closed-source models, examining their performance in both isolated and joint tasks of veracity prediction and explanation generation. Importantly, we employ a dual evaluation approach comprising previously established automatic metrics and a novel set of criteria through human evaluation. Our automatic evaluation indicates that, within the zero-shot scenario, GPT-4 emerges as the standout performer, but in few-shot and parameter-efficient fine-tuning contexts, open-source models demonstrate their capacity to not only bridge the performance gap but, in some instances, surpass GPT-4. Human evaluation reveals yet more nuance as well as indicating potential problems with the gold explanations.", }

This paper presents a comprehensive analysis of explainable fact-checking through a series of experiments, focusing on the ability of large language models to verify public health claims and provide explanations or justifications for their veracity assessments. We examine the effectiveness of zero/few-shot prompting and parameter-efficient fine-tuning across various open and closed-source models, examining their performance in both isolated and joint tasks of veracity prediction and explanation generation. Importantly, we employ a dual evaluation approach comprising previously established automatic metrics and a novel set of criteria through human evaluation. Our automatic evaluation indicates that, within the zero-shot scenario, GPT-4 emerges as the standout performer, but in few-shot and parameter-efficient fine-tuning contexts, open-source models demonstrate their capacity to not only bridge the performance gap but, in some instances, surpass GPT-4. Human evaluation reveals yet more nuance as well as indicating potential problems with the gold explanations.

Tell Me Why: Explainable Public Health Fact-Checking with Large Language Models

trustnlp-1.21

Poster

2405.09454

https://aclanthology.org/2024.vardial-1.1.bib

https://aclanthology.org/2024.vardial-1.1/

@inproceedings{chifu-etal-2024-vardial, title = "{V}ar{D}ial Evaluation Campaign 2024: Commonsense Reasoning in Dialects and Multi-Label Similar Language Identification", author = "Chifu, Adrian-Gabriel and Glava{\v{s}}, Goran and Ionescu, Radu Tudor and Ljube{\v{s}}i{\'c}, Nikola and Mileti{\'c}, Aleksandra and Mileti{\'c}, Filip and Scherrer, Yves and Vuli{\'c}, Ivan", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.1", doi = "10.18653/v1/2024.vardial-1.1", pages = "1--15", abstract = "This report presents the results of the shared tasks organized as part of the VarDial Evaluation Campaign 2024. The campaign is part of the eleventh workshop on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects (VarDial), co-located with NAACL 2024. Two shared tasks were included this year: dialectal causal commonsense reasoning (DIALECT-COPA), and Multi-label classification of similar languages (DSL-ML). Both tasks were organized for the first time this year, but DSL-ML partially overlaps with the DSL-TL task organized in 2023.", }

This report presents the results of the shared tasks organized as part of the VarDial Evaluation Campaign 2024. The campaign is part of the eleventh workshop on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects (VarDial), co-located with NAACL 2024. Two shared tasks were included this year: dialectal causal commonsense reasoning (DIALECT-COPA), and Multi-label classification of similar languages (DSL-ML). Both tasks were organized for the first time this year, but DSL-ML partially overlaps with the DSL-TL task organized in 2023.

VarDial Evaluation Campaign 2024: Commonsense Reasoning in Dialects and Multi-Label Similar Language Identification

vardial-1.1

Poster

https://aclanthology.org/2024.vardial-1.2.bib

https://aclanthology.org/2024.vardial-1.2/

@inproceedings{bagheri-nezhad-agrawal-2024-drives, title = "What Drives Performance in Multilingual Language Models?", author = "Bagheri Nezhad, Sina and Agrawal, Ameeta", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.2", doi = "10.18653/v1/2024.vardial-1.2", pages = "16--27", abstract = "This study investigates the factors influencing the performance of multilingual large language models (MLLMs) across diverse languages. We study 6 MLLMs, including masked language models, autoregressive models, and instruction-tuned LLMs, on the SIB-200 dataset, a topic classification dataset encompassing 204 languages. Our analysis considers three scenarios: ALL languages, SEEN languages (present in the model{'}s pretraining data), and UNSEEN languages (not present or documented in the model{'}s pretraining data in any meaningful way). We examine the impact of factors such as pretraining data size, general resource availability, language family, and script type on model performance. Decision tree analysis reveals that pretraining data size is the most influential factor for SEEN languages. However, interestingly, script type and language family become more crucial for UNSEEN languages, highlighting the importance of cross-lingual transfer learning. Notably, model size and architecture do not significantly alter the most important features identified. Our findings provide valuable insights into the strengths and limitations of current MLLMs and hope to guide the development of more effective and equitable multilingual NLP systems.", }

This study investigates the factors influencing the performance of multilingual large language models (MLLMs) across diverse languages. We study 6 MLLMs, including masked language models, autoregressive models, and instruction-tuned LLMs, on the SIB-200 dataset, a topic classification dataset encompassing 204 languages. Our analysis considers three scenarios: ALL languages, SEEN languages (present in the model{'}s pretraining data), and UNSEEN languages (not present or documented in the model{'}s pretraining data in any meaningful way). We examine the impact of factors such as pretraining data size, general resource availability, language family, and script type on model performance. Decision tree analysis reveals that pretraining data size is the most influential factor for SEEN languages. However, interestingly, script type and language family become more crucial for UNSEEN languages, highlighting the importance of cross-lingual transfer learning. Notably, model size and architecture do not significantly alter the most important features identified. Our findings provide valuable insights into the strengths and limitations of current MLLMs and hope to guide the development of more effective and equitable multilingual NLP systems.

What Drives Performance in Multilingual Language Models?

vardial-1.2

Poster

2404.19159

https://aclanthology.org/2024.vardial-1.3.bib

https://aclanthology.org/2024.vardial-1.3/

@inproceedings{dolev-etal-2024-whisper, title = "Does Whisper Understand {S}wiss {G}erman? An Automatic, Qualitative, and Human Evaluation", author = {Dolev, Eyal and Lutz, Clemens and Aepli, No{\"e}mi}, editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.3", doi = "10.18653/v1/2024.vardial-1.3", pages = "28--40", abstract = {Whisper is a state-of-the-art automatic speech recognition (ASR) model (Radford et al., 2022). Although Swiss German dialects are allegedly not part of Whisper{'}s training data, preliminary experiments showed Whisper can transcribe Swiss German quite well, with the output being a speech translation into Standard German. To gain a better understanding of Whisper{'}s performance on Swiss German, we systematically evaluate it using automatic, qualitative, and human evaluation. We test its performance on three existing test sets: SwissDial (Dogan-Sch{\"o}nberger et al., 2021), STT4SG-350 (Pl{\"u}ss et al., 2023), and Swiss Parliaments Corpus (Pl{\"u}ss et al., 2021). In addition, we create a new test set for this study based on short mock clinical interviews. To automatically evaluate performance, we used word error rate (WER) and BLEU. We also conducted a qualitative analysis of Whisper{'}s performance, discussing its strengths and weaknesses. Finally, 28 people participated in a survey evaluating Whisper{'}s performance. All of our evaluations showed that Whisper is a viable ASR system for Swiss German, so long as the Standard German output is desired.}, }

Whisper is a state-of-the-art automatic speech recognition (ASR) model (Radford et al., 2022). Although Swiss German dialects are allegedly not part of Whisper{'}s training data, preliminary experiments showed Whisper can transcribe Swiss German quite well, with the output being a speech translation into Standard German. To gain a better understanding of Whisper{'}s performance on Swiss German, we systematically evaluate it using automatic, qualitative, and human evaluation. We test its performance on three existing test sets: SwissDial (Dogan-Sch{\"o}nberger et al., 2021), STT4SG-350 (Pl{\"u}ss et al., 2023), and Swiss Parliaments Corpus (Pl{\"u}ss et al., 2021). In addition, we create a new test set for this study based on short mock clinical interviews. To automatically evaluate performance, we used word error rate (WER) and BLEU. We also conducted a qualitative analysis of Whisper{'}s performance, discussing its strengths and weaknesses. Finally, 28 people participated in a survey evaluating Whisper{'}s performance. All of our evaluations showed that Whisper is a viable ASR system for Swiss German, so long as the Standard German output is desired.

Does Whisper Understand Swiss German? An Automatic, Qualitative, and Human Evaluation

vardial-1.3

Poster

2404.19310

https://aclanthology.org/2024.vardial-1.4.bib

https://aclanthology.org/2024.vardial-1.4/

@inproceedings{mohamed-eida-etal-2024-well, title = "How Well Do Tweets Represent Sub-Dialects of {E}gyptian {A}rabic?", author = "Mohamed Eida, Mai and Nassar, Mayar and Dunn, Jonathan", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.4", doi = "10.18653/v1/2024.vardial-1.4", pages = "41--55", abstract = "How well does naturally-occurring digital text, such as Tweets, represent sub-dialects of Egyptian Arabic (EA)? This paper focuses on two EA sub-dialects: Cairene Egyptian Arabic (CEA) and Sa{'}idi Egyptian Arabic (SEA). We use morphological markers from ground-truth dialect surveys as a distance measure across four geo-referenced datasets. Results show that CEA markers are prevalent as expected in CEA geo-referenced tweets, while SEA markers are limited across SEA geo-referenced tweets. SEA tweets instead show a prevalence of CEA markers and higher usage of Modern Standard Arabic. We conclude that corpora intended to represent sub-dialects of EA do not accurately represent sub-dialects outside of the Cairene variety. This finding calls into question the validity of relying on tweets alone to represent dialectal differences.", }

How well does naturally-occurring digital text, such as Tweets, represent sub-dialects of Egyptian Arabic (EA)? This paper focuses on two EA sub-dialects: Cairene Egyptian Arabic (CEA) and Sa{'}idi Egyptian Arabic (SEA). We use morphological markers from ground-truth dialect surveys as a distance measure across four geo-referenced datasets. Results show that CEA markers are prevalent as expected in CEA geo-referenced tweets, while SEA markers are limited across SEA geo-referenced tweets. SEA tweets instead show a prevalence of CEA markers and higher usage of Modern Standard Arabic. We conclude that corpora intended to represent sub-dialects of EA do not accurately represent sub-dialects outside of the Cairene variety. This finding calls into question the validity of relying on tweets alone to represent dialectal differences.

How Well Do Tweets Represent Sub-Dialects of Egyptian Arabic?

vardial-1.4

Poster

https://aclanthology.org/2024.vardial-1.5.bib

https://aclanthology.org/2024.vardial-1.5/

@inproceedings{espana-bonet-etal-2024-elote, title = "When Elote, Choclo and Mazorca are not the Same. Isomorphism-Based Perspective to the {S}panish Varieties Divergences", author = "Espa{\~n}a-Bonet, Cristina and Bhatt, Ankur and Dutta Chowdhury, Koel and Barr{\'o}n-Cede{\~n}o, Alberto", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.5", doi = "10.18653/v1/2024.vardial-1.5", pages = "56--77", abstract = "Spanish is an official language in 20 countries; in 19 of them, it arrived by means of overseas colonisation. Its close contact with several coexistent languages and the rich regional and cultural diversity has produced varieties which divert from each other. We study these divergences in a data-based approach and according to their qualitative and quantitative effects in word embeddings. We generate embeddings for Spanish in 24 countries and examine the topology of the spaces. Due to the similarities between varieties {---}in contrast to what happens to different languages in bilingual topological studies{---} we first scrutinise the behaviour of three isomorphism measures in (quasi-)isomorphic settings: relational similarity, Eigenvalue similarity and Gromov-Hausdorff distance. We then use the most trustworthy measure to quantify the divergences among varieties. Finally, we use the departures from isomorphism to build relational trees for the Spanish varieties by hierarchical clustering.", }

Spanish is an official language in 20 countries; in 19 of them, it arrived by means of overseas colonisation. Its close contact with several coexistent languages and the rich regional and cultural diversity has produced varieties which divert from each other. We study these divergences in a data-based approach and according to their qualitative and quantitative effects in word embeddings. We generate embeddings for Spanish in 24 countries and examine the topology of the spaces. Due to the similarities between varieties {---}in contrast to what happens to different languages in bilingual topological studies{---} we first scrutinise the behaviour of three isomorphism measures in (quasi-)isomorphic settings: relational similarity, Eigenvalue similarity and Gromov-Hausdorff distance. We then use the most trustworthy measure to quantify the divergences among varieties. Finally, we use the departures from isomorphism to build relational trees for the Spanish varieties by hierarchical clustering.

When Elote, Choclo and Mazorca are not the Same. Isomorphism-Based Perspective to the Spanish Varieties Divergences

vardial-1.5

Poster

https://aclanthology.org/2024.vardial-1.6.bib

https://aclanthology.org/2024.vardial-1.6/

@inproceedings{hopton-aepli-2024-modeling, title = "Modeling Orthographic Variation in {O}ccitan{'}s Dialects", author = {Hopton, Zachary and Aepli, No{\"e}mi}, editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.6", doi = "10.18653/v1/2024.vardial-1.6", pages = "78--88", abstract = "Effectively normalizing spellings in textual data poses a considerable challenge, especially for low-resource languages lacking standardized writing systems. In this study, we fine-tuned a multilingual model with data from several Occitan dialects and conducted a series of experiments to assess the model{'}s representations of these dialects. For evaluation purposes, we compiled a parallel lexicon encompassing four Occitan dialects.Intrinsic evaluations of the model{'}s embeddings revealed that surface similarity between the dialects strengthened representations. When the model was further fine-tuned for part-of-speech tagging, its performance was robust to dialectical variation, even when trained solely on part-of-speech data from a single dialect. Our findings suggest that large multilingual models minimize the need for spelling normalization during pre-processing.", }

Effectively normalizing spellings in textual data poses a considerable challenge, especially for low-resource languages lacking standardized writing systems. In this study, we fine-tuned a multilingual model with data from several Occitan dialects and conducted a series of experiments to assess the model{'}s representations of these dialects. For evaluation purposes, we compiled a parallel lexicon encompassing four Occitan dialects.Intrinsic evaluations of the model{'}s embeddings revealed that surface similarity between the dialects strengthened representations. When the model was further fine-tuned for part-of-speech tagging, its performance was robust to dialectical variation, even when trained solely on part-of-speech data from a single dialect. Our findings suggest that large multilingual models minimize the need for spelling normalization during pre-processing.

Modeling Orthographic Variation in Occitan's Dialects

vardial-1.6

Poster

2404.19315

https://aclanthology.org/2024.vardial-1.7.bib

https://aclanthology.org/2024.vardial-1.7/

@inproceedings{ljubesic-etal-2024-dialect, title = "{DIALECT}-{COPA}: Extending the Standard Translations of the {COPA} Causal Commonsense Reasoning Dataset to {S}outh {S}lavic Dialects", author = "Ljube{\v{s}}i{\'c}, Nikola and Galant, Nada and Ben{\v{c}}ina, Sonja and {\v{C}}ibej, Jaka and Milosavljevi{\'c}, Stefan and Rupnik, Peter and Kuzman, Taja", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.7", doi = "10.18653/v1/2024.vardial-1.7", pages = "89--98", abstract = "The paper presents new causal commonsense reasoning datasets for South Slavic dialects, based on the Choice of Plausible Alternatives (COPA) dataset. The dialectal datasets are built by translating by native dialect speakers from the English original and the corresponding standard translation. Three dialects are covered {--} the Cerkno dialect of Slovenian, the Chakavian dialect of Croatian and the Torlak dialect of Serbian. The datasets are the first resource for evaluation of large language models on South Slavic dialects, as well as among the first commonsense reasoning datasets on dialects overall. The paper describes specific challenges met during the translation process. A comparison of the dialectal datasets with their standard language counterparts shows a varying level of character-level, word-level and lexicon-level deviation of dialectal text from the standard datasets. The observed differences are well reproduced in initial zero-shot and 10-shot experiments, where the Slovenian Cerkno dialect and the Croatian Chakavian dialect show significantly lower results than the Torlak dialect. These results show also for the dialectal datasets to be significantly more challenging than the standard datasets. Finally, in-context learning on just 10 examples shows to improve the results dramatically, especially for the dialects with the lowest results.", }

The paper presents new causal commonsense reasoning datasets for South Slavic dialects, based on the Choice of Plausible Alternatives (COPA) dataset. The dialectal datasets are built by translating by native dialect speakers from the English original and the corresponding standard translation. Three dialects are covered {--} the Cerkno dialect of Slovenian, the Chakavian dialect of Croatian and the Torlak dialect of Serbian. The datasets are the first resource for evaluation of large language models on South Slavic dialects, as well as among the first commonsense reasoning datasets on dialects overall. The paper describes specific challenges met during the translation process. A comparison of the dialectal datasets with their standard language counterparts shows a varying level of character-level, word-level and lexicon-level deviation of dialectal text from the standard datasets. The observed differences are well reproduced in initial zero-shot and 10-shot experiments, where the Slovenian Cerkno dialect and the Croatian Chakavian dialect show significantly lower results than the Torlak dialect. These results show also for the dialectal datasets to be significantly more challenging than the standard datasets. Finally, in-context learning on just 10 examples shows to improve the results dramatically, especially for the dialects with the lowest results.

DIALECT-COPA: Extending the Standard Translations of the COPA Causal Commonsense Reasoning Dataset to South Slavic Dialects

vardial-1.7

Poster

https://aclanthology.org/2024.vardial-1.8.bib

https://aclanthology.org/2024.vardial-1.8/

@inproceedings{muller-etal-2024-role, title = "The Role of Adverbs in Language Variety Identification: The Case of {P}ortuguese Multi-Word Adverbs", author = {M{\"u}ller, Izabela and Mamede, Nuno and Baptista, Jorge}, editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.8", doi = "10.18653/v1/2024.vardial-1.8", pages = "99--106", abstract = "This paper aims to assess the role of multiword compound adverbs in distinguishing Brazilian Portuguese (PT-BR) from European Portuguese (PT-PT). Two key factors underpin this focus: Firstly, multiword expressions often provide less ambiguity compared to single words, even when their meaning is idiomatic (non-compositional). Secondly, despite constituting a significant portion of lexicons in many languages, they are frequently overlooked in Natural Language Processing, possibly due to their heterogeneous nature and lexical range.For this study, a large lexicon of Portuguese multiword adverbs (3,665) annotated with diatopic information regarding language variety was utilized. The paper investigates the distribution of this category in a corpus consisting in excerpts from journalistic texts sourced from the DSL (Dialect and Similar Language) corpus, representing Brazilian (PT-BR) and European Portuguese (PT-PT), respectively, each partition containing 18,000 sentences.Results indicate a substantial similarity between the two varieties, with a considerable overlap in the lexicon of multiword adverbs. Additionally, specific adverbs unique to each language variety were identified. Lexical entries recognized in the corpus represent 18.2{\%} (PT-BR) to 19.5{\%} (PT-PT) of the lexicon, and approximately 5,700 matches in each partition. While many of the matches are spurious due to ambiguity with otherwise non-idiomatic, free strings, occurrences of adverbs marked as exclusive to one variety in texts from the other variety are rare.", }

This paper aims to assess the role of multiword compound adverbs in distinguishing Brazilian Portuguese (PT-BR) from European Portuguese (PT-PT). Two key factors underpin this focus: Firstly, multiword expressions often provide less ambiguity compared to single words, even when their meaning is idiomatic (non-compositional). Secondly, despite constituting a significant portion of lexicons in many languages, they are frequently overlooked in Natural Language Processing, possibly due to their heterogeneous nature and lexical range.For this study, a large lexicon of Portuguese multiword adverbs (3,665) annotated with diatopic information regarding language variety was utilized. The paper investigates the distribution of this category in a corpus consisting in excerpts from journalistic texts sourced from the DSL (Dialect and Similar Language) corpus, representing Brazilian (PT-BR) and European Portuguese (PT-PT), respectively, each partition containing 18,000 sentences.Results indicate a substantial similarity between the two varieties, with a considerable overlap in the lexicon of multiword adverbs. Additionally, specific adverbs unique to each language variety were identified. Lexical entries recognized in the corpus represent 18.2{\%} (PT-BR) to 19.5{\%} (PT-PT) of the lexicon, and approximately 5,700 matches in each partition. While many of the matches are spurious due to ambiguity with otherwise non-idiomatic, free strings, occurrences of adverbs marked as exclusive to one variety in texts from the other variety are rare.

The Role of Adverbs in Language Variety Identification: The Case of Portuguese Multi-Word Adverbs

vardial-1.8

Poster

https://aclanthology.org/2024.vardial-1.9.bib

https://aclanthology.org/2024.vardial-1.9/

@inproceedings{maehlum-scherrer-2024-nomusic, title = "{N}o{M}usic - The {N}orwegian Multi-Dialectal Slot and Intent Detection Corpus", author = "M{\ae}hlum, Petter and Scherrer, Yves", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.9", doi = "10.18653/v1/2024.vardial-1.9", pages = "107--116", abstract = "This paper presents a new textual resource for Norwegian and its dialects. The NoMusic corpus contains Norwegian translations of the xSID dataset, an evaluation dataset for spoken language understanding (slot and intent detection). The translations cover Norwegian Bokm{\aa}l, as well as eight dialects from three of the four major Norwegian dialect areas. To our knowledge, this is the first multi-parallel resource for written Norwegian dialects, and the first evaluation dataset for slot and intent detection focusing on non-standard Norwegian varieties. In this paper, we describe the annotation process and provide some analyses on the types of linguistic variation that can be found in the dataset.", }

This paper presents a new textual resource for Norwegian and its dialects. The NoMusic corpus contains Norwegian translations of the xSID dataset, an evaluation dataset for spoken language understanding (slot and intent detection). The translations cover Norwegian Bokm{\aa}l, as well as eight dialects from three of the four major Norwegian dialect areas. To our knowledge, this is the first multi-parallel resource for written Norwegian dialects, and the first evaluation dataset for slot and intent detection focusing on non-standard Norwegian varieties. In this paper, we describe the annotation process and provide some analyses on the types of linguistic variation that can be found in the dataset.

NoMusic - The Norwegian Multi-Dialectal Slot and Intent Detection Corpus

vardial-1.9

Poster

https://aclanthology.org/2024.vardial-1.10.bib

https://aclanthology.org/2024.vardial-1.10/

@inproceedings{olabisi-agrawal-2024-understanding, title = "Understanding Position Bias Effects on Fairness in Social Multi-Document Summarization", author = "Olabisi, Olubusayo and Agrawal, Ameeta", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.10", doi = "10.18653/v1/2024.vardial-1.10", pages = "117--129", abstract = "Text summarization models have typically focused on optimizing aspects of quality such as fluency, relevance, and coherence, particularly in the context of news articles. However, summarization models are increasingly being used to summarize diverse sources of text, such as social media data, that encompass a wide demographic user base. It is thus crucial to assess not only the quality of the generated summaries, but also the extent to which they can fairly represent the opinions of diverse social groups. Position bias, a long-known issue in news summarization, has received limited attention in the context of social multi-document summarization. We deeply investigate this phenomenon by analyzing the effect of group ordering in input documents when summarizing tweets from three distinct linguistic communities: African-American English, Hispanic-aligned Language, and White-aligned Language. Our empirical analysis shows that although the textual quality of the summaries remains consistent regardless of the input document order, in terms of fairness, the results vary significantly depending on how the dialect groups are presented in the input data. Our results suggest that position bias manifests differently in social multi-document summarization, severely impacting the fairness of summarization models.", }

Text summarization models have typically focused on optimizing aspects of quality such as fluency, relevance, and coherence, particularly in the context of news articles. However, summarization models are increasingly being used to summarize diverse sources of text, such as social media data, that encompass a wide demographic user base. It is thus crucial to assess not only the quality of the generated summaries, but also the extent to which they can fairly represent the opinions of diverse social groups. Position bias, a long-known issue in news summarization, has received limited attention in the context of social multi-document summarization. We deeply investigate this phenomenon by analyzing the effect of group ordering in input documents when summarizing tweets from three distinct linguistic communities: African-American English, Hispanic-aligned Language, and White-aligned Language. Our empirical analysis shows that although the textual quality of the summaries remains consistent regardless of the input document order, in terms of fairness, the results vary significantly depending on how the dialect groups are presented in the input data. Our results suggest that position bias manifests differently in social multi-document summarization, severely impacting the fairness of summarization models.

Understanding Position Bias Effects on Fairness in Social Multi-Document Summarization

vardial-1.10

Poster

2405.01790

https://huggingface.co/papers/2405.01790

https://aclanthology.org/2024.vardial-1.11.bib

https://aclanthology.org/2024.vardial-1.11/

@inproceedings{ondrejova-suppa-2024-llms, title = "Can {LLM}s Handle Low-Resource Dialects? A Case Study on Translation and Common Sense Reasoning in {\v{S}}ari{\v{s}}", author = "Ondrejov{\'a}, Vikt{\'o}ria and {\v{S}}uppa, Marek", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.11", doi = "10.18653/v1/2024.vardial-1.11", pages = "130--139", abstract = "While Large Language Models (LLMs) have demonstrated considerable potential in advancing natural language processing in dialect-specific contexts, their effectiveness in these settings has yet to be thoroughly assessed. This study introduces a case study on {\v{S}}ari{\v{s}}, a dialect of Slovak, which is itself a language with fewer resources, focusing on Machine Translation and Common Sense Reasoning tasks. We employ LLMs in a zero-shot configuration and for data augmentation to refine Slovak-{\v{S}}ari{\v{s}} and {\v{S}}ari{\v{s}}-Slovak translation models. The accuracy of these models is then manually verified by native speakers. Additionally, we introduce {\v{S}}ari{\v{s}}COPA, a new dataset for causal common sense reasoning, which, alongside SlovakCOPA, serves to evaluate LLM{'}s performance in a zero-shot framework. Our findings highlight LLM{'}s capabilities in processing low-resource dialects and suggest a viable approach for initiating dialect-specific translation models in such contexts.", }

While Large Language Models (LLMs) have demonstrated considerable potential in advancing natural language processing in dialect-specific contexts, their effectiveness in these settings has yet to be thoroughly assessed. This study introduces a case study on {\v{S}}ari{\v{s}}, a dialect of Slovak, which is itself a language with fewer resources, focusing on Machine Translation and Common Sense Reasoning tasks. We employ LLMs in a zero-shot configuration and for data augmentation to refine Slovak-{\v{S}}ari{\v{s}} and {\v{S}}ari{\v{s}}-Slovak translation models. The accuracy of these models is then manually verified by native speakers. Additionally, we introduce {\v{S}}ari{\v{s}}COPA, a new dataset for causal common sense reasoning, which, alongside SlovakCOPA, serves to evaluate LLM{'}s performance in a zero-shot framework. Our findings highlight LLM{'}s capabilities in processing low-resource dialects and suggest a viable approach for initiating dialect-specific translation models in such contexts.

Can LLMs Handle Low-Resource Dialects? A Case Study on Translation and Common Sense Reasoning in Šariš

vardial-1.11

Poster

https://aclanthology.org/2024.vardial-1.12.bib

https://aclanthology.org/2024.vardial-1.12/

@inproceedings{pugh-tyers-2024-experiments, title = "Experiments in Multi-Variant Natural Language Processing for {N}ahuatl", author = "Pugh, Robert and Tyers, Francis", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.12", doi = "10.18653/v1/2024.vardial-1.12", pages = "140--151", abstract = "Linguistic variation is a complicating factor for digital language technologies. This is particularly true for languages that lack an official {``}standard{''} variety, including many regional and minoritized languages. In this paper, we describe a set of experiments focused on multivariant natural language processing for the Nahuatl, an indigenous Mexican language with a high level of linguistic variation and no single recognized standard variant. Using small (10k tokens), recently-published annotated datasets for two Nahuatl variants, we compare the performance of single-variant, cross-variant, and joint training, and explore how different models perform on a third Nahuatl variant, unseen in training. These results and the subsequent discussion contribute to efforts of developing low-resource NLP that is robust to diatopic variation. We share all code used to process the data and run the experiments.", }

Linguistic variation is a complicating factor for digital language technologies. This is particularly true for languages that lack an official {``}standard{''} variety, including many regional and minoritized languages. In this paper, we describe a set of experiments focused on multivariant natural language processing for the Nahuatl, an indigenous Mexican language with a high level of linguistic variation and no single recognized standard variant. Using small (10k tokens), recently-published annotated datasets for two Nahuatl variants, we compare the performance of single-variant, cross-variant, and joint training, and explore how different models perform on a third Nahuatl variant, unseen in training. These results and the subsequent discussion contribute to efforts of developing low-resource NLP that is robust to diatopic variation. We share all code used to process the data and run the experiments.

Experiments in Multi-Variant Natural Language Processing for Nahuatl

vardial-1.12

Poster

https://aclanthology.org/2024.vardial-1.13.bib

https://aclanthology.org/2024.vardial-1.13/

@inproceedings{simons-etal-2024-highly, title = "Highly Granular Dialect Normalization and Phonological Dialect Translation for {L}imburgish", author = "Simons, Andreas and De Pascale, Stefano and Franco, Karlien", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.13", doi = "10.18653/v1/2024.vardial-1.13", pages = "152--162", abstract = "We study highly granular dialect normalization and phonological dialect translation on Limburgish, a non-standardized low-resource language with a wide variation in spelling conventions and phonology. We find improvements to the traditional transformer by embedding the geographic coordinates of dialects in dialect normalization tasks and use these geographically-embedded transformers to translate words between the phonologies of different dialects. These results are found to be consistent with notions in traditional Limburgish dialectology.", }

We study highly granular dialect normalization and phonological dialect translation on Limburgish, a non-standardized low-resource language with a wide variation in spelling conventions and phonology. We find improvements to the traditional transformer by embedding the geographic coordinates of dialects in dialect normalization tasks and use these geographically-embedded transformers to translate words between the phonologies of different dialects. These results are found to be consistent with notions in traditional Limburgish dialectology.

Highly Granular Dialect Normalization and Phonological Dialect Translation for Limburgish

vardial-1.13

Poster

https://aclanthology.org/2024.vardial-1.14.bib

https://aclanthology.org/2024.vardial-1.14/

@inproceedings{sterner-2024-multilingual, title = "Multilingual Identification of {E}nglish Code-Switching", author = "Sterner, Igor", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.14", doi = "10.18653/v1/2024.vardial-1.14", pages = "163--173", abstract = "Code-switching research depends on fine-grained language identification. In this work, we study existing corpora used to train token-level language identification systems. We aggregate these corpora with a consistent labelling scheme and train a system to identify English code-switching in multilingual text. We show that the system identifies code-switching in unseen language pairs with absolute measure 2.3-4.6{\%} better than language-pair-specific SoTA. We also analyse the correlation between typological similarity of the languages and difficulty in recognizing code-switching.", }

Code-switching research depends on fine-grained language identification. In this work, we study existing corpora used to train token-level language identification systems. We aggregate these corpora with a consistent labelling scheme and train a system to identify English code-switching in multilingual text. We show that the system identifies code-switching in unseen language pairs with absolute measure 2.3-4.6{\%} better than language-pair-specific SoTA. We also analyse the correlation between typological similarity of the languages and difficulty in recognizing code-switching.

Multilingual Identification of English Code-Switching

vardial-1.14

Poster

https://aclanthology.org/2024.vardial-1.15.bib

https://aclanthology.org/2024.vardial-1.15/

@inproceedings{verkijk-etal-2024-studying, title = "Studying Language Variation Considering the Re-Usability of Modern Theories, Tools and Resources for Annotating Explicit and Implicit Events in Centuries Old Text", author = "Verkijk, Stella and Sommerauer, Pia and Vossen, Piek", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.15", doi = "10.18653/v1/2024.vardial-1.15", pages = "174--187", abstract = "This paper discusses the re-usibility of existing approaches, tools and automatic techniques for the annotation and detection of events in a challenging variant of centuries old Dutch written in the archives of the Dutch East India Company. We describe our annotation process and provide a thorough analysis of different versions of manually annotated data and the first automatic results from two fine-tuned Language Models. Through the analysis of this complete process, the paper studies two things: to what extent we can use NLP theories and tasks formulated for modern English to formulate an annotation task for Early Modern Dutch and to what extent we can use NLP models and tools built for modern Dutch (and other languages) on Early Modern Dutch. We believe these analyses give us insight into how to deal with the large variation language showcases in describing events, and how this variation may differ accross domains. We release the annotation guidelines, annotated data, and code.", }

This paper discusses the re-usibility of existing approaches, tools and automatic techniques for the annotation and detection of events in a challenging variant of centuries old Dutch written in the archives of the Dutch East India Company. We describe our annotation process and provide a thorough analysis of different versions of manually annotated data and the first automatic results from two fine-tuned Language Models. Through the analysis of this complete process, the paper studies two things: to what extent we can use NLP theories and tasks formulated for modern English to formulate an annotation task for Early Modern Dutch and to what extent we can use NLP models and tools built for modern Dutch (and other languages) on Early Modern Dutch. We believe these analyses give us insight into how to deal with the large variation language showcases in describing events, and how this variation may differ accross domains. We release the annotation guidelines, annotated data, and code.

Studying Language Variation Considering the Re-Usability of Modern Theories, Tools and Resources for Annotating Explicit and Implicit Events in Centuries Old Text

vardial-1.15

Poster

https://aclanthology.org/2024.vardial-1.16.bib

https://aclanthology.org/2024.vardial-1.16/

@inproceedings{vicente-cheng-2024-language, title = "Language Identification of {P}hilippine Creole {S}panish: Discriminating {C}havacano From Related Languages", author = "Vicente, Aileen Joan and Cheng, Charibeth", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.16", doi = "10.18653/v1/2024.vardial-1.16", pages = "188--196", abstract = "Chavacano is a Spanish Creole widely spoken in the southern regions of the Philippines. It is one of the many Philippine languages yet to be studied computationally. This paper presents the development of a language identification model of Chavacano to distinguish it from languages that influence its creolization using character convolutional networks. Unlike studies that discriminated similar languages based on geographical proximity, this paper reports a similarity focused on the creolization of a language. We established the similarity of Chavacano and its related languages, Spanish, Portuguese, Cebuano, and Hiligaynon, from the number of common words in the corpus for all languages. We report an accuracy of 93{\%} for the model generated using ten filters with a filter width of 5. The training experiments reveal that increasing the filter width, number of filters, or training epochs is unnecessary even if the accuracy increases because the generated models present irregular learning behavior or may have already been overfitted. This study also demonstrates that the character features extracted from convolutional neural networks, similar to n-grams, are sufficient in identifying Chavacano. Future work on the language identification of Chavacano includes improving classification accuracy for short or code-switched texts for practical applications such as social media sensors for disaster response and management.", }

Chavacano is a Spanish Creole widely spoken in the southern regions of the Philippines. It is one of the many Philippine languages yet to be studied computationally. This paper presents the development of a language identification model of Chavacano to distinguish it from languages that influence its creolization using character convolutional networks. Unlike studies that discriminated similar languages based on geographical proximity, this paper reports a similarity focused on the creolization of a language. We established the similarity of Chavacano and its related languages, Spanish, Portuguese, Cebuano, and Hiligaynon, from the number of common words in the corpus for all languages. We report an accuracy of 93{\%} for the model generated using ten filters with a filter width of 5. The training experiments reveal that increasing the filter width, number of filters, or training epochs is unnecessary even if the accuracy increases because the generated models present irregular learning behavior or may have already been overfitted. This study also demonstrates that the character features extracted from convolutional neural networks, similar to n-grams, are sufficient in identifying Chavacano. Future work on the language identification of Chavacano includes improving classification accuracy for short or code-switched texts for practical applications such as social media sensors for disaster response and management.

Language Identification of Philippine Creole Spanish: Discriminating Chavacano From Related Languages

vardial-1.16

Poster

https://aclanthology.org/2024.vardial-1.17.bib

https://aclanthology.org/2024.vardial-1.17/

@inproceedings{faisal-anastasopoulos-2024-data, title = "Data-Augmentation-Based Dialectal Adaptation for {LLM}s", author = "Faisal, Fahim and Anastasopoulos, Antonios", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.17", doi = "10.18653/v1/2024.vardial-1.17", pages = "197--208", abstract = "This report presents gmnlp{'}s participation to the Dialect-Copa shared task at VarDial 2024 (Chifu et al., 2024), which focuses on evaluating the commonsense reasoning capabilities of large language models (LLMs) on South Slavic micro-dialects. The task aims to assess how well LLMs can handle non-standard dialectal varieties, as their performance on standard languages is already well-established. We propose an approach that combines the strengths of different types of language models and leverages data augmentation techniques to improve task performance on three South Slavic dialects: Chakavian, Cherkano, and Torlak. We conduct experiments using a language-family-focused encoder-based model (BERTi{\'c}) and a domain-agnostic multilingual model (AYA-101). Our results demonstrate that the proposed data augmentation techniques lead to substantial performance gains across all three test datasets in the open-source model category. This work highlights the practical utility of data augmentation and the potential of LLMs in handling non-standard dialectal varieties, contributing to the broader goal of advancing natural language understanding in low-resource and dialectal settings.", }

This report presents gmnlp{'}s participation to the Dialect-Copa shared task at VarDial 2024 (Chifu et al., 2024), which focuses on evaluating the commonsense reasoning capabilities of large language models (LLMs) on South Slavic micro-dialects. The task aims to assess how well LLMs can handle non-standard dialectal varieties, as their performance on standard languages is already well-established. We propose an approach that combines the strengths of different types of language models and leverages data augmentation techniques to improve task performance on three South Slavic dialects: Chakavian, Cherkano, and Torlak. We conduct experiments using a language-family-focused encoder-based model (BERTi{\'c}) and a domain-agnostic multilingual model (AYA-101). Our results demonstrate that the proposed data augmentation techniques lead to substantial performance gains across all three test datasets in the open-source model category. This work highlights the practical utility of data augmentation and the potential of LLMs in handling non-standard dialectal varieties, contributing to the broader goal of advancing natural language understanding in low-resource and dialectal settings.

Data-Augmentation-Based Dialectal Adaptation for LLMs

vardial-1.17

Poster

2404.08092

https://aclanthology.org/2024.vardial-1.18.bib

https://aclanthology.org/2024.vardial-1.18/

@inproceedings{ljubesic-etal-2024-jsi, title = {{JSI} and {W}{\"u}{NLP} at the {DIALECT}-{COPA} Shared Task: In-Context Learning From Just a Few Dialectal Examples Gets You Quite Far}, author = "Ljube{\v{s}}i{\'c}, Nikola and Kuzman, Taja and Rupnik, Peter and Vuli{\'c}, Ivan and Schmidt, Fabian and Glava{\v{s}}, Goran", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.18", doi = "10.18653/v1/2024.vardial-1.18", pages = "209--219", abstract = {The paper presents the JSI and W{\"u}NLP systems submitted to the DIALECT-COPA shared task on causal commonsense reasoning in dialectal texts. Jointly, we compare LLM-based zero-shot and few-shot in-context inference (JSI team), and task-specific few-shot fine-tuning, in English and respective standard language, with zero-shot cross-lingual transfer (ZS-XLT) to the test dialects (W{\"u}NLP team). Given the very strong zero-shot and especially few-shot in-context learning (ICL) performance, we further investigate whether task semantics, or language/dialect semantics explain the strong performance, showing that a significant part of the improvement indeed stems from learning the language or dialect semantics from the in-context examples, with only a minor contribution from understanding the nature of the task. The higher importance of the dialect semantics to the task semantics is further shown by the finding that the in-context learning with only a few dialectal instances achieves comparable results to the supervised fine-tuning approach on hundreds of instances in standard language.}, }

The paper presents the JSI and W{\"u}NLP systems submitted to the DIALECT-COPA shared task on causal commonsense reasoning in dialectal texts. Jointly, we compare LLM-based zero-shot and few-shot in-context inference (JSI team), and task-specific few-shot fine-tuning, in English and respective standard language, with zero-shot cross-lingual transfer (ZS-XLT) to the test dialects (W{\"u}NLP team). Given the very strong zero-shot and especially few-shot in-context learning (ICL) performance, we further investigate whether task semantics, or language/dialect semantics explain the strong performance, showing that a significant part of the improvement indeed stems from learning the language or dialect semantics from the in-context examples, with only a minor contribution from understanding the nature of the task. The higher importance of the dialect semantics to the task semantics is further shown by the finding that the in-context learning with only a few dialectal instances achieves comparable results to the supervised fine-tuning approach on hundreds of instances in standard language.

JSI and WüNLP at the DIALECT-COPA Shared Task: In-Context Learning From Just a Few Dialectal Examples Gets You Quite Far

vardial-1.18

Poster

https://aclanthology.org/2024.vardial-1.19.bib

https://aclanthology.org/2024.vardial-1.19/

@inproceedings{perak-etal-2024-incorporating, title = "Incorporating Dialect Understanding Into {LLM} Using {RAG} and Prompt Engineering Techniques for Causal Commonsense Reasoning", author = "Perak, Benedikt and Beliga, Slobodan and Me{\v{s}}trovi{\'c}, Ana", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.19", doi = "10.18653/v1/2024.vardial-1.19", pages = "220--229", abstract = "The choice of plausible alternatives (COPA) task requires selecting the most plausible outcome from two choices based on understanding the causal relationships presented in a given text.This paper outlines several approaches and model adaptation strategies to the VarDial 2024 DIALECT-COPA shared task, focusing on causal commonsense reasoning in South-Slavic dialects. We utilize and evaluate the GPT-4 model in combination with various prompts engineering and the Retrieval-Augmented Generation (RAG) technique. Initially, we test and compare the performance of GPT-4 with simple and advanced prompts on the COPA task across three dialects: Cerkno, Chakavian and Torlak. Next, we enhance prompts using the RAG technique specifically for the Chakavian and Cerkno dialect. This involves creating an extended Chakavian-English and Cerkno-Slovene lexical dictionary and integrating it into the prompts. Our findings indicate that the most complex approach, which combines an advanced prompt with an injected dictionary, yields the highest performance on the DIALECT-COPA task.", }

The choice of plausible alternatives (COPA) task requires selecting the most plausible outcome from two choices based on understanding the causal relationships presented in a given text.This paper outlines several approaches and model adaptation strategies to the VarDial 2024 DIALECT-COPA shared task, focusing on causal commonsense reasoning in South-Slavic dialects. We utilize and evaluate the GPT-4 model in combination with various prompts engineering and the Retrieval-Augmented Generation (RAG) technique. Initially, we test and compare the performance of GPT-4 with simple and advanced prompts on the COPA task across three dialects: Cerkno, Chakavian and Torlak. Next, we enhance prompts using the RAG technique specifically for the Chakavian and Cerkno dialect. This involves creating an extended Chakavian-English and Cerkno-Slovene lexical dictionary and integrating it into the prompts. Our findings indicate that the most complex approach, which combines an advanced prompt with an injected dictionary, yields the highest performance on the DIALECT-COPA task.

Incorporating Dialect Understanding Into LLM Using RAG and Prompt Engineering Techniques for Causal Commonsense Reasoning

vardial-1.19

Poster

https://aclanthology.org/2024.vardial-1.20.bib

https://aclanthology.org/2024.vardial-1.20/

@inproceedings{gillin-2024-one, title = "One-Shot Prompt for Language Variety Identification", author = "Gillin, Nat", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.20", doi = "10.18653/v1/2024.vardial-1.20", pages = "230--234", abstract = "We present a one-shot prompting approach to multi-class classification for similar language identification with off-the-shelf pre-trained large language model that is not particularly trained or tuned for the language identification task. Without post-training or fine-tuning the model, we simply include one example per class when prompting the model and surprisingly the model to generate the language andlocale labels accordingly.", }

We present a one-shot prompting approach to multi-class classification for similar language identification with off-the-shelf pre-trained large language model that is not particularly trained or tuned for the language identification task. Without post-training or fine-tuning the model, we simply include one example per class when prompting the model and surprisingly the model to generate the language andlocale labels accordingly.

One-Shot Prompt for Language Variety Identification

vardial-1.20

Poster

https://aclanthology.org/2024.vardial-1.21.bib

https://aclanthology.org/2024.vardial-1.21/

@inproceedings{ngo-etal-2024-improving, title = "Improving Multi-Label Classification of Similar Languages by Semantics-Aware Word Embeddings", author = "Ngo, The and Nguyen, Thi Anh and Ha, My and Nguyen, Thi Minh and Le-Hong, Phuong", editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.21", doi = "10.18653/v1/2024.vardial-1.21", pages = "235--240", abstract = "The VLP team participated in the DSL-ML shared task of the VarDial 2024 workshop which aims to distinguish texts in similar languages. This paper presents our approach to solving the problem and discusses our experimental and official results. We propose to integrate semantics-aware word embeddings which are learned from ConceptNet into a bidirectional long short-term memory network. This approach achieves good performance {--} our sys- tem is ranked in the top two or three of the best performing teams for the task.", }

The VLP team participated in the DSL-ML shared task of the VarDial 2024 workshop which aims to distinguish texts in similar languages. This paper presents our approach to solving the problem and discusses our experimental and official results. We propose to integrate semantics-aware word embeddings which are learned from ConceptNet into a bidirectional long short-term memory network. This approach achieves good performance {--} our sys- tem is ranked in the top two or three of the best performing teams for the task.

Improving Multi-Label Classification of Similar Languages by Semantics-Aware Word Embeddings

vardial-1.21

Poster

https://aclanthology.org/2024.vardial-1.22.bib

https://aclanthology.org/2024.vardial-1.22/

@inproceedings{saleva-palen-michel-2024-brandeis, title = "{B}randeis at {V}ar{D}ial 2024 {DSL}-{ML} Shared Task: Multilingual Models, Simple Baselines and Data Augmentation", author = {S{\"a}lev{\"a}, Jonne and Palen-Michel, Chester}, editor = {Scherrer, Yves and Jauhiainen, Tommi and Ljube{\v{s}}i{\'c}, Nikola and Zampieri, Marcos and Nakov, Preslav and Tiedemann, J{\"o}rg}, booktitle = "Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.vardial-1.22", doi = "10.18653/v1/2024.vardial-1.22", pages = "241--251", abstract = "This paper describes the Brandeis University submission to VarDial 2024 DSL-ML Shared Task on multilabel classification for discriminating between similar languages. Our submission consists of three entries per language to the closed track, where no additional data was permitted. Our approach involves a set of simple non-neural baselines using logistic regression, random forests and support vector machines. We follow this by experimenting with finetuning multilingual BERT, either on a single language or all the languages concatenated together.In addition to benchmarking the model architectures against one another on the development set, we perform extensive hyperparameter tuning, which is afforded by the small size of the training data.Our experiments on the development set suggest that finetuned mBERT systems significantly benefit most languages compared to the baseline.However, on the test set, our results indicate that simple models based on scikit-learn can perform surprisingly well and even outperform pretrained language models, as we see with BCMS.Our submissions achieve the best performance on all languages as reported by the organizers. Except for Spanish and French, our non-neural baseline also ranks in the top 3 for all other languages.", }

This paper describes the Brandeis University submission to VarDial 2024 DSL-ML Shared Task on multilabel classification for discriminating between similar languages. Our submission consists of three entries per language to the closed track, where no additional data was permitted. Our approach involves a set of simple non-neural baselines using logistic regression, random forests and support vector machines. We follow this by experimenting with finetuning multilingual BERT, either on a single language or all the languages concatenated together.In addition to benchmarking the model architectures against one another on the development set, we perform extensive hyperparameter tuning, which is afforded by the small size of the training data.Our experiments on the development set suggest that finetuned mBERT systems significantly benefit most languages compared to the baseline.However, on the test set, our results indicate that simple models based on scikit-learn can perform surprisingly well and even outperform pretrained language models, as we see with BCMS.Our submissions achieve the best performance on all languages as reported by the organizers. Except for Spanish and French, our non-neural baseline also ranks in the top 3 for all other languages.

Brandeis at VarDial 2024 DSL-ML Shared Task: Multilingual Models, Simple Baselines and Data Augmentation

vardial-1.22

Poster

https://aclanthology.org/2024.woah-1.1.bib

https://aclanthology.org/2024.woah-1.1/

@inproceedings{de-kock-hovy-2024-investigating, title = "Investigating radicalisation indicators in online extremist communities", author = "De Kock, Christine and Hovy, Eduard", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.1", doi = "10.18653/v1/2024.woah-1.1", pages = "1--12", abstract = "We identify and analyse three sociolinguistic indicators of radicalisation within online extremist forums: hostility, longevity and social connectivity. We develop models to predict the maximum degree of each indicator measured over an individual{'}s lifetime, based on a minimal number of initial interactions. Drawing on data from two diverse extremist communities, our results demonstrate that NLP methods are effective at prioritising at-risk users. This work offers practical insights for intervention strategies and policy development, and highlights an important but under-studied research direction.", }

We identify and analyse three sociolinguistic indicators of radicalisation within online extremist forums: hostility, longevity and social connectivity. We develop models to predict the maximum degree of each indicator measured over an individual{'}s lifetime, based on a minimal number of initial interactions. Drawing on data from two diverse extremist communities, our results demonstrate that NLP methods are effective at prioritising at-risk users. This work offers practical insights for intervention strategies and policy development, and highlights an important but under-studied research direction.

Investigating radicalisation indicators in online extremist communities

woah-1.1

Poster

https://aclanthology.org/2024.woah-1.2.bib

https://aclanthology.org/2024.woah-1.2/

@inproceedings{pustet-etal-2024-detection, title = "Detection of Conspiracy Theories Beyond Keyword Bias in {G}erman-Language Telegram Using Large Language Models", author = "Pustet, Milena and Steffen, Elisabeth and Mihaljevic, Helena", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.2", doi = "10.18653/v1/2024.woah-1.2", pages = "13--27", abstract = "To protect users from massive hateful content, existing works studied automated hate speech detection. Despite the existing efforts, one question remains: do automated hate speech detectors conform to social media content policies? A platform{'}s content policies are a checklist of content moderated by the social media platform. Because content moderation rules are often uniquely defined, existing hate speech datasets cannot directly answer this question. This work seeks to answer this question by creating HateModerate, a dataset for testing the behaviors of automated content moderators against content policies. First, we engage 28 annotators and GPT in a six-step annotation process, resulting in a list of hateful and non-hateful test suites matching each of Facebook{'}s 41 hate speech policies. Second, we test the performance of state-of-the-art hate speech detectors against HateModerate, revealing substantial failures these models have in their conformity to the policies. Third, using HateModerate, we augment the training data of a top-downloaded hate detector on HuggingFace. We observe significant improvement in the models{'} conformity to content policies while having comparable scores on the original test data. Our dataset and code can be found in the attachment.", }

To protect users from massive hateful content, existing works studied automated hate speech detection. Despite the existing efforts, one question remains: do automated hate speech detectors conform to social media content policies? A platform{'}s content policies are a checklist of content moderated by the social media platform. Because content moderation rules are often uniquely defined, existing hate speech datasets cannot directly answer this question. This work seeks to answer this question by creating HateModerate, a dataset for testing the behaviors of automated content moderators against content policies. First, we engage 28 annotators and GPT in a six-step annotation process, resulting in a list of hateful and non-hateful test suites matching each of Facebook{'}s 41 hate speech policies. Second, we test the performance of state-of-the-art hate speech detectors against HateModerate, revealing substantial failures these models have in their conformity to the policies. Third, using HateModerate, we augment the training data of a top-downloaded hate detector on HuggingFace. We observe significant improvement in the models{'} conformity to content policies while having comparable scores on the original test data. Our dataset and code can be found in the attachment.

Detection of Conspiracy Theories Beyond Keyword Bias in German-Language Telegram Using Large Language Models

woah-1.2

Poster

2404.17985

https://aclanthology.org/2024.woah-1.3.bib

https://aclanthology.org/2024.woah-1.3/

@inproceedings{ilevbare-etal-2024-ekohate, title = "{E}ko{H}ate: Abusive Language and Hate Speech Detection for Code-switched Political Discussions on {N}igerian {T}witter", author = "Ilevbare, Comfort and Alabi, Jesujoba and Adelani, David Ifeoluwa and Bakare, Firdous and Abiola, Oluwatoyin and Adeyemo, Oluwaseyi", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.3", doi = "10.18653/v1/2024.woah-1.3", pages = "28--37", abstract = "Nigerians have a notable online presence and actively discuss political and topical matters. This was particularly evident throughout the 2023 general election, where Twitter was used for campaigning, fact-checking and verification, and even positive and negative discourse. However, little or none has been done in the detection of abusive language and hate speech in Nigeria. In this paper, we curated code-switched Twitter data directed at three musketeers of the governorship election on the most populous and economically vibrant state in Nigeria; Lagos state, with the view to detect offensive speech in political discussions. We developed EkoHate{---}an abusive language and hate speech dataset for political discussions between the three candidates and their followers using a binary (normal vs offensive) and fine-grained four-label annotation scheme. We analysed our dataset and provided an empirical evaluation of state-of-the-art methods across both supervised and cross-lingual transfer learning settings. In the supervised setting, our evaluation results in both binary and four-label annotation schemes show that we can achieve 95.1 and 70.3 F1 points respectively. Furthermore, we show that our dataset adequately transfers very well to three publicly available offensive datasets (OLID, HateUS2020, and FountaHate), generalizing to political discussions in other regions like the US.", }

Nigerians have a notable online presence and actively discuss political and topical matters. This was particularly evident throughout the 2023 general election, where Twitter was used for campaigning, fact-checking and verification, and even positive and negative discourse. However, little or none has been done in the detection of abusive language and hate speech in Nigeria. In this paper, we curated code-switched Twitter data directed at three musketeers of the governorship election on the most populous and economically vibrant state in Nigeria; Lagos state, with the view to detect offensive speech in political discussions. We developed EkoHate{---}an abusive language and hate speech dataset for political discussions between the three candidates and their followers using a binary (normal vs offensive) and fine-grained four-label annotation scheme. We analysed our dataset and provided an empirical evaluation of state-of-the-art methods across both supervised and cross-lingual transfer learning settings. In the supervised setting, our evaluation results in both binary and four-label annotation schemes show that we can achieve 95.1 and 70.3 F1 points respectively. Furthermore, we show that our dataset adequately transfers very well to three publicly available offensive datasets (OLID, HateUS2020, and FountaHate), generalizing to political discussions in other regions like the US.

EkoHate: Abusive Language and Hate Speech Detection for Code-switched Political Discussions on Nigerian Twitter

woah-1.3

Poster

2404.18180

https://aclanthology.org/2024.woah-1.4.bib

https://aclanthology.org/2024.woah-1.4/

@inproceedings{zhang-etal-2024-study-class, title = "A Study of the Class Imbalance Problem in Abusive Language Detection", author = "Zhang, Yaqi and Hangya, Viktor and Fraser, Alexander", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.4", doi = "10.18653/v1/2024.woah-1.4", pages = "38--51", abstract = "Abusive language detection has drawn increasing interest in recent years. However, a less systematically explored obstacle is label imbalance, i.e., the amount of abusive data is much lower than non-abusive data, leading to performance issues. The aim of this work is to conduct a comprehensive comparative study of popular methods for addressing the class imbalance issue. We explore 10 well-known approaches on 8 datasets with distinct characteristics: binary or multi-class, moderately or largely imbalanced, focusing on various types of abuse, etc. Additionally, we pro-pose two novel methods specialized for abuse detection: AbusiveLexiconAug and ExternalDataAug, which enrich the training data using abusive lexicons and external abusive datasets, respectively. We conclude that: 1) our AbusiveLexiconAug approach, random oversampling, and focal loss are the most versatile methods on various datasets; 2) focal loss tends to yield peak model performance; 3) oversampling and focal loss provide promising results for binary datasets and small multi-class sets, while undersampling and weighted cross-entropy are more suitable for large multi-class sets; 4) most methods are sensitive to hyperparameters, yet our suggested choice of hyperparameters provides a good starting point.", }

Abusive language detection has drawn increasing interest in recent years. However, a less systematically explored obstacle is label imbalance, i.e., the amount of abusive data is much lower than non-abusive data, leading to performance issues. The aim of this work is to conduct a comprehensive comparative study of popular methods for addressing the class imbalance issue. We explore 10 well-known approaches on 8 datasets with distinct characteristics: binary or multi-class, moderately or largely imbalanced, focusing on various types of abuse, etc. Additionally, we pro-pose two novel methods specialized for abuse detection: AbusiveLexiconAug and ExternalDataAug, which enrich the training data using abusive lexicons and external abusive datasets, respectively. We conclude that: 1) our AbusiveLexiconAug approach, random oversampling, and focal loss are the most versatile methods on various datasets; 2) focal loss tends to yield peak model performance; 3) oversampling and focal loss provide promising results for binary datasets and small multi-class sets, while undersampling and weighted cross-entropy are more suitable for large multi-class sets; 4) most methods are sensitive to hyperparameters, yet our suggested choice of hyperparameters provides a good starting point.

A Study of the Class Imbalance Problem in Abusive Language Detection

woah-1.4

Poster

https://aclanthology.org/2024.woah-1.5.bib

https://aclanthology.org/2024.woah-1.5/

@inproceedings{vargas-etal-2024-hausahate, title = "{H}ausa{H}ate: An Expert Annotated Corpus for {H}ausa Hate Speech Detection", author = "Vargas, Francielle and Guimar{\~a}es, Samuel and Muhammad, Shamsuddeen Hassan and Alves, Diego and Ahmad, Ibrahim Said and Abdulmumin, Idris and Mohamed, Diallo and Pardo, Thiago and Benevenuto, Fabr{\'\i}cio", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.5", doi = "10.18653/v1/2024.woah-1.5", pages = "52--58", abstract = "We introduce the first expert annotated corpus of Facebook comments for Hausa hate speech detection. The corpus titled HausaHate comprises 2,000 comments extracted from Western African Facebook pages and manually annotated by three Hausa native speakers, who are also NLP experts. Our corpus was annotated using two different layers. We first labeled each comment according to a binary classification: offensive versus non-offensive. Then, offensive comments were also labeled according to hate speech targets: race, gender and none. Lastly, a baseline model using fine-tuned LLM for Hausa hate speech detection is presented, highlighting the challenges of hate speech detection tasks for indigenous languages in Africa, as well as future advances.", }

We introduce the first expert annotated corpus of Facebook comments for Hausa hate speech detection. The corpus titled HausaHate comprises 2,000 comments extracted from Western African Facebook pages and manually annotated by three Hausa native speakers, who are also NLP experts. Our corpus was annotated using two different layers. We first labeled each comment according to a binary classification: offensive versus non-offensive. Then, offensive comments were also labeled according to hate speech targets: race, gender and none. Lastly, a baseline model using fine-tuned LLM for Hausa hate speech detection is presented, highlighting the challenges of hate speech detection tasks for indigenous languages in Africa, as well as future advances.

HausaHate: An Expert Annotated Corpus for Hausa Hate Speech Detection

woah-1.5

Poster

https://aclanthology.org/2024.woah-1.6.bib

https://aclanthology.org/2024.woah-1.6/

@inproceedings{anastasi-etal-2024-vida, title = "{VIDA}: The Visual Incel Data Archive. A Theory-oriented Annotated Dataset To Enhance Hate Detection Through Visual Culture", author = "Anastasi, Selenia and Schneider, Florian and Biemann, Chris and Fischer, Tim", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.6", doi = "10.18653/v1/2024.woah-1.6", pages = "59--67", abstract = "Images increasingly constitute a larger portion of internet content, encoding even more complex meanings. Recent studies have highlight the pivotal role of visual communication in the spread of extremist content, particularly that associated with right-wing political ideologies. However, the capability of machine learning systems to recognize such meanings, sometimes implicit, remains limited. To enable future research in this area, we introduce and release VIDA, the Visual Incel Data Archive, a multimodal dataset comprising visual material and internet memes collected from two main Incel communities (Italian and Anglophone) known for their extremist misogynistic content. Following the analytical framework of Shifman (2014), we propose a new taxonomy for annotation across three main levels of analysis: content, form, and stance (hate). This allows for the association of images with fine-grained contextual information that help to identify the presence of offensiveness and a broader set of cultural references, enhancing the understanding of more nuanced aspects in visual communication. In this work we present a statistical analysis of the annotated dataset as well as discuss annotation examples and future line of research.", }

Images increasingly constitute a larger portion of internet content, encoding even more complex meanings. Recent studies have highlight the pivotal role of visual communication in the spread of extremist content, particularly that associated with right-wing political ideologies. However, the capability of machine learning systems to recognize such meanings, sometimes implicit, remains limited. To enable future research in this area, we introduce and release VIDA, the Visual Incel Data Archive, a multimodal dataset comprising visual material and internet memes collected from two main Incel communities (Italian and Anglophone) known for their extremist misogynistic content. Following the analytical framework of Shifman (2014), we propose a new taxonomy for annotation across three main levels of analysis: content, form, and stance (hate). This allows for the association of images with fine-grained contextual information that help to identify the presence of offensiveness and a broader set of cultural references, enhancing the understanding of more nuanced aspects in visual communication. In this work we present a statistical analysis of the annotated dataset as well as discuss annotation examples and future line of research.

VIDA: The Visual Incel Data Archive. A Theory-oriented Annotated Dataset To Enhance Hate Detection Through Visual Culture

woah-1.6

Poster

https://aclanthology.org/2024.woah-1.7.bib

https://aclanthology.org/2024.woah-1.7/

@inproceedings{omrani-etal-2024-towards, title = "Towards a Unified Framework for Adaptable Problematic Content Detection via Continual Learning", author = "Omrani, Ali and Salkhordeh Ziabari, Alireza and Golazizian, Preni and Sorensen, Jeffrey and Dehghani, Morteza", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.7", doi = "10.18653/v1/2024.woah-1.7", pages = "68--109", abstract = "Detecting problematic content, such as hate speech, is a multifaceted and ever-changing task, influenced by social dynamics, user populations, diversity of sources, and evolving language. There has been significant efforts, both in academia and in industry, to develop annotated resources that capture various aspects of problematic content. Due to researchers{'} diverse objectives, these annotations are often inconsistent and hence, reports of progress on the detection of problematic content are fragmented. This pattern is expected to persist unless we pool these resources, taking into account the dynamic nature of this issue. In this paper, we propose integrating the available resources, leveraging their dynamic nature to break this pattern, and introduce a continual learning framework and benchmark for problematic content detection. Our benchmark, comprising 84 related tasks, creates a novel measure of progress: prioritizing the adaptability of classifiers to evolving tasks over excelling in specific tasks. To ensure continuous relevance, our benchmark is designed for seamless integration of new tasks. Our results demonstrate that continual learning methods outperform static approaches by up to 17{\%} and 4{\%} AUC in capturing the evolving content and adapting to novel forms of problematic content", }

Detecting problematic content, such as hate speech, is a multifaceted and ever-changing task, influenced by social dynamics, user populations, diversity of sources, and evolving language. There has been significant efforts, both in academia and in industry, to develop annotated resources that capture various aspects of problematic content. Due to researchers{'} diverse objectives, these annotations are often inconsistent and hence, reports of progress on the detection of problematic content are fragmented. This pattern is expected to persist unless we pool these resources, taking into account the dynamic nature of this issue. In this paper, we propose integrating the available resources, leveraging their dynamic nature to break this pattern, and introduce a continual learning framework and benchmark for problematic content detection. Our benchmark, comprising 84 related tasks, creates a novel measure of progress: prioritizing the adaptability of classifiers to evolving tasks over excelling in specific tasks. To ensure continuous relevance, our benchmark is designed for seamless integration of new tasks. Our results demonstrate that continual learning methods outperform static approaches by up to 17{\%} and 4{\%} AUC in capturing the evolving content and adapting to novel forms of problematic content

Towards a Unified Framework for Adaptable Problematic Content Detection via Continual Learning

woah-1.7

Poster

2309.16905

https://aclanthology.org/2024.woah-1.8.bib

https://aclanthology.org/2024.woah-1.8/

@inproceedings{liebeskind-etal-2024-linguistics, title = "From Linguistics to Practice: a Case Study of Offensive Language Taxonomy in {H}ebrew", author = "Liebeskind, Chaya and Litvak, Marina and Vanetik, Natalia", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.8", doi = "10.18653/v1/2024.woah-1.8", pages = "110--117", abstract = "The perception of offensive language varies based on cultural, social, and individual perspectives. With the spread of social media, there has been an increase in offensive content online, necessitating advanced solutions for its identification and moderation. This paper addresses the practical application of an offensive language taxonomy, specifically targeting Hebrew social media texts. By introducing a newly annotated dataset, modeled after the taxonomy of explicit offensive language of (Lewandowska-Tomaszczyk et al., 2023)„ we provide a comprehensive examination of various degrees and aspects of offensive language. Our findings indicate the complexities involved in the classification of such content. We also outline the implications of relying on fixed taxonomies for Hebrew.", }

The perception of offensive language varies based on cultural, social, and individual perspectives. With the spread of social media, there has been an increase in offensive content online, necessitating advanced solutions for its identification and moderation. This paper addresses the practical application of an offensive language taxonomy, specifically targeting Hebrew social media texts. By introducing a newly annotated dataset, modeled after the taxonomy of explicit offensive language of (Lewandowska-Tomaszczyk et al., 2023)„ we provide a comprehensive examination of various degrees and aspects of offensive language. Our findings indicate the complexities involved in the classification of such content. We also outline the implications of relying on fixed taxonomies for Hebrew.

From Linguistics to Practice: a Case Study of Offensive Language Taxonomy in Hebrew

woah-1.8

Poster

https://aclanthology.org/2024.woah-1.9.bib

https://aclanthology.org/2024.woah-1.9/

@inproceedings{lislevand-etal-2024-estimating, title = "Estimating the Emotion of Disgust in {G}reek Parliament Records", author = "Lislevand, Vanessa and Pavlopoulos, John and Louridas, Panos and Dritsa, Konstantina", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.9", doi = "10.18653/v1/2024.woah-1.9", pages = "118--135", abstract = "We present an analysis of the sentiment in Greek political speech, by focusing on the most frequently occurring emotion in electoral data, the emotion of {``}disgust{''}. We show that emotion classification is generally tough, but high accuracy can be achieved for that particular emotion. Using our best-performing model to classify political records of the Greek Parliament Corpus from 1989 to 2020, we studied the points in time when this emotion was frequently occurring and we ranked the Greek political parties based on their estimated score. We then devised an algorithm to investigate the emotional context shift of words that describe specific conditions and that can be used to stigmatise. Given that early detection of such word usage is essential for policy-making, we report two words we found being increasingly used in a negative emotional context, and one that is likely to be carrying stigma, in the studied parliamentary records.", }

We present an analysis of the sentiment in Greek political speech, by focusing on the most frequently occurring emotion in electoral data, the emotion of {``}disgust{''}. We show that emotion classification is generally tough, but high accuracy can be achieved for that particular emotion. Using our best-performing model to classify political records of the Greek Parliament Corpus from 1989 to 2020, we studied the points in time when this emotion was frequently occurring and we ranked the Greek political parties based on their estimated score. We then devised an algorithm to investigate the emotional context shift of words that describe specific conditions and that can be used to stigmatise. Given that early detection of such word usage is essential for policy-making, we report two words we found being increasingly used in a negative emotional context, and one that is likely to be carrying stigma, in the studied parliamentary records.

Estimating the Emotion of Disgust in Greek Parliament Records

woah-1.9

Poster

https://aclanthology.org/2024.woah-1.10.bib

https://aclanthology.org/2024.woah-1.10/

@inproceedings{fillies-paschke-2024-simple, title = "Simple {LLM} based Approach to Counter Algospeak", author = "Fillies, Jan and Paschke, Adrian", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.10", doi = "10.18653/v1/2024.woah-1.10", pages = "136--145", abstract = "With the use of algorithmic moderation on online communication platforms, an increase in adaptive language aiming to evade the automatic detection of problematic content has been observed. One form of this adapted language is known as {``}Algospeak{''} and is most commonly associated with large social media platforms, e.g., TikTok. It builds upon Leetspeak or online slang with its explicit intention to avoid machine readability. The machine-learning algorithms employed to automate the process of content moderation mostly rely on human-annotated datasets and supervised learning, often not adjusted for a wide variety of languages and changes in language. This work uses linguistic examples identified in research literature to introduce a taxonomy for Algospeak and shows that with the use of an LLM (GPT-4), 79.4{\%} of the established terms can be corrected to their true form, or if needed, their underlying associated concepts. With an example sentence, 98.5{\%} of terms are correctly identified. This research demonstrates that LLMs are the future in solving the current problem of moderation avoidance by Algospeak.", }

With the use of algorithmic moderation on online communication platforms, an increase in adaptive language aiming to evade the automatic detection of problematic content has been observed. One form of this adapted language is known as {``}Algospeak{''} and is most commonly associated with large social media platforms, e.g., TikTok. It builds upon Leetspeak or online slang with its explicit intention to avoid machine readability. The machine-learning algorithms employed to automate the process of content moderation mostly rely on human-annotated datasets and supervised learning, often not adjusted for a wide variety of languages and changes in language. This work uses linguistic examples identified in research literature to introduce a taxonomy for Algospeak and shows that with the use of an LLM (GPT-4), 79.4{\%} of the established terms can be corrected to their true form, or if needed, their underlying associated concepts. With an example sentence, 98.5{\%} of terms are correctly identified. This research demonstrates that LLMs are the future in solving the current problem of moderation avoidance by Algospeak.

Simple LLM based Approach to Counter Algospeak

woah-1.10

Poster

https://aclanthology.org/2024.woah-1.11.bib

https://aclanthology.org/2024.woah-1.11/

@inproceedings{ashraf-etal-2024-harnessing, title = "Harnessing Personalization Methods to Identify and Predict Unreliable Information Spreader Behavior", author = "Ashraf, Shaina and Gruschka, Fabio and Flek, Lucie and Welch, Charles", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.11", doi = "10.18653/v1/2024.woah-1.11", pages = "146--158", abstract = "Studies on detecting and understanding the spread of unreliable news on social media have identified key characteristic differences between reliable and unreliable posts. These differences in language use also vary in expression across individuals, making it important to consider personal factors in unreliable news detection. The application of personalization methods for this has been made possible by recent publication of datasets with user histories, though this area is still largely unexplored. In this paper we present approaches to represent social media users in order to improve performance on three tasks: (1) classification of unreliable news posts, (2) classification of unreliable news spreaders, and, (3) prediction of the spread of unreliable news. We compare the User2Vec method from previous work to two other approaches; a learnable user embedding layer trained with the downstream task, and a representation derived from an authorship attribution classifier. We demonstrate that the implemented strategies substantially improve classification performance over state-of-the-art and provide initial results on the task of unreliable news prediction.", }

Studies on detecting and understanding the spread of unreliable news on social media have identified key characteristic differences between reliable and unreliable posts. These differences in language use also vary in expression across individuals, making it important to consider personal factors in unreliable news detection. The application of personalization methods for this has been made possible by recent publication of datasets with user histories, though this area is still largely unexplored. In this paper we present approaches to represent social media users in order to improve performance on three tasks: (1) classification of unreliable news posts, (2) classification of unreliable news spreaders, and, (3) prediction of the spread of unreliable news. We compare the User2Vec method from previous work to two other approaches; a learnable user embedding layer trained with the downstream task, and a representation derived from an authorship attribution classifier. We demonstrate that the implemented strategies substantially improve classification performance over state-of-the-art and provide initial results on the task of unreliable news prediction.

Harnessing Personalization Methods to Identify and Predict Unreliable Information Spreader Behavior

woah-1.11

Poster

https://aclanthology.org/2024.woah-1.12.bib

https://aclanthology.org/2024.woah-1.12/

@inproceedings{kim-etal-2024-robust, title = "Robust Safety Classifier Against Jailbreaking Attacks: Adversarial Prompt Shield", author = "Kim, Jinhwa and Derakhshan, Ali and Harris, Ian", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.12", doi = "10.18653/v1/2024.woah-1.12", pages = "159--170", abstract = "Large Language Models{'} safety remains a critical concern due to their vulnerability to jailbreaking attacks, which can prompt these systems to produce harmful and malicious responses. Safety classifiers, computational models trained to discern and mitigate potentially harmful, offensive, or unethical outputs, offer a practical solution to address this issue. However, despite their potential, existing safety classifiers often fail when exposed to adversarial attacks such as gradient-optimized suffix attacks. In response, our study introduces Adversarial Prompt Shield (APS), a lightweight safety classifier model that excels in detection accuracy and demonstrates resilience against unseen jailbreaking prompts. We also introduce efficiently generated adversarial training datasets, named Bot Adversarial Noisy Dialogue (BAND), which are designed to fortify the classifier{'}s robustness. Through extensive testing on various safety tasks and unseen jailbreaking attacks, we demonstrate the effectiveness and resilience of our models. Evaluations show that our classifier has the potential to significantly reduce the Attack Success Rate by up to 44.9{\%}. This advance paves the way for the next generation of more reliable and resilient Large Language Models.", }

Large Language Models{'} safety remains a critical concern due to their vulnerability to jailbreaking attacks, which can prompt these systems to produce harmful and malicious responses. Safety classifiers, computational models trained to discern and mitigate potentially harmful, offensive, or unethical outputs, offer a practical solution to address this issue. However, despite their potential, existing safety classifiers often fail when exposed to adversarial attacks such as gradient-optimized suffix attacks. In response, our study introduces Adversarial Prompt Shield (APS), a lightweight safety classifier model that excels in detection accuracy and demonstrates resilience against unseen jailbreaking prompts. We also introduce efficiently generated adversarial training datasets, named Bot Adversarial Noisy Dialogue (BAND), which are designed to fortify the classifier{'}s robustness. Through extensive testing on various safety tasks and unseen jailbreaking attacks, we demonstrate the effectiveness and resilience of our models. Evaluations show that our classifier has the potential to significantly reduce the Attack Success Rate by up to 44.9{\%}. This advance paves the way for the next generation of more reliable and resilient Large Language Models.

Robust Safety Classifier Against Jailbreaking Attacks: Adversarial Prompt Shield

woah-1.12

Poster

https://aclanthology.org/2024.woah-1.13.bib

https://aclanthology.org/2024.woah-1.13/

@inproceedings{reyes-ramirez-etal-2024-improving, title = "Improving aggressiveness detection using a data augmentation technique based on a Diffusion Language Model", author = "Reyes-Ram{\'\i}rez, Antonio and Arag{\'o}n, Mario and S{\'a}nchez-Vega, Fernando and L{\'o}pez-Monroy, Adrian", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.13", doi = "10.18653/v1/2024.woah-1.13", pages = "171--177", abstract = "Cyberbullying has grown in recent years, largely attributed to the proliferation of social media users. This phenomenon manifests in various forms, such as hate speech and offensive language, increasing the necessity of effective detection models to tackle this problem. Most approaches focus on supervised algorithms, which have an important drawback{---}they heavily depend on the availability of ample training data. This paper attempts to tackle this insufficient data problem using data augmentation (DA) techniques. Concretely, we propose a novel data augmentation technique based on a Diffusion Language Model (DLA). We compare our proposed method against well-known DA techniques, such as contextual augmentation and Easy Data Augmentation (EDA). Our findings reveal a slight but promising improvement, leading to more robust results with very low variance. Additionally, we provide a comprehensive qualitative analysis using classification errors, and complementary analysis, shedding light on the nuances of our approach.", }

Cyberbullying has grown in recent years, largely attributed to the proliferation of social media users. This phenomenon manifests in various forms, such as hate speech and offensive language, increasing the necessity of effective detection models to tackle this problem. Most approaches focus on supervised algorithms, which have an important drawback{---}they heavily depend on the availability of ample training data. This paper attempts to tackle this insufficient data problem using data augmentation (DA) techniques. Concretely, we propose a novel data augmentation technique based on a Diffusion Language Model (DLA). We compare our proposed method against well-known DA techniques, such as contextual augmentation and Easy Data Augmentation (EDA). Our findings reveal a slight but promising improvement, leading to more robust results with very low variance. Additionally, we provide a comprehensive qualitative analysis using classification errors, and complementary analysis, shedding light on the nuances of our approach.

Improving aggressiveness detection using a data augmentation technique based on a Diffusion Language Model

woah-1.13

Poster

https://aclanthology.org/2024.woah-1.14.bib

https://aclanthology.org/2024.woah-1.14/

@inproceedings{andersen-etal-2024-mexican, title = "The {M}exican Gayze: A Computational Analysis of the Attitudes towards the {LGBT}+ Population in {M}exico on Social Media Across a Decade", author = "Andersen, Scott and Ojeda-Trueba, Segio-Luis and V{\'a}squez, Juan and Bel-Enguix, Gemma", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.14", doi = "10.18653/v1/2024.woah-1.14", pages = "178--200", abstract = "Thanks to the popularity of social media, data generated by online communities provides an abundant source of diverse language information. This abundance of data allows NLP practitioners and computational linguists to analyze sociolinguistic phenomena occurring in digital communication. In this paper, we analyze the Twitter discourse around the Mexican Spanish-speaking LGBT+ community. For this, we evaluate how the polarity of some nouns related to the LGBT+ community has evolved in conversational settings using a corpus of tweets that cover a time span of ten years. We hypothesize that social media{'}s fast-moving, turbulent linguistic environment encourages language evolution faster than ever before. Our results indicate that most of the inspected terms have undergone some shift in denotation or connotation. No other generalizations can be observed in the data, given the difficulty that current NLP methods have to account for polysemy, and the wide differences between the various subgroups that make up the LGBT+ community. A fine-grained analysis of a series of LGBT+-related lexical terms is also included in this work.", }

Thanks to the popularity of social media, data generated by online communities provides an abundant source of diverse language information. This abundance of data allows NLP practitioners and computational linguists to analyze sociolinguistic phenomena occurring in digital communication. In this paper, we analyze the Twitter discourse around the Mexican Spanish-speaking LGBT+ community. For this, we evaluate how the polarity of some nouns related to the LGBT+ community has evolved in conversational settings using a corpus of tweets that cover a time span of ten years. We hypothesize that social media{'}s fast-moving, turbulent linguistic environment encourages language evolution faster than ever before. Our results indicate that most of the inspected terms have undergone some shift in denotation or connotation. No other generalizations can be observed in the data, given the difficulty that current NLP methods have to account for polysemy, and the wide differences between the various subgroups that make up the LGBT+ community. A fine-grained analysis of a series of LGBT+-related lexical terms is also included in this work.

The Mexican Gayze: A Computational Analysis of the Attitudes towards the LGBT+ Population in Mexico on Social Media Across a Decade

woah-1.14

Poster

https://aclanthology.org/2024.woah-1.15.bib

https://aclanthology.org/2024.woah-1.15/

@inproceedings{arun-etal-2024-x, title = "{X}-posing Free Speech: Examining the Impact of Moderation Relaxation on Online Social Networks", author = "Arun, Arvindh and Chhatani, Saurav and An, Jisun and Kumaraguru, Ponnurangam", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.15", doi = "10.18653/v1/2024.woah-1.15", pages = "201--211", abstract = "We investigate the impact of free speech and the relaxation of moderation on online social media platforms using Elon Musk{'}s takeover of Twitter as a case study. By curating a dataset of over 10 million tweets, our study employs a novel framework combining content and network analysis. Our findings reveal a significant increase in the distribution of certain forms of hate content, particularly targeting the LGBTQ+ community and liberals. Network analysis reveals the formation of cohesive hate communities facilitated by influential bridge users, with substantial growth in interactions hinting at increased hate production and diffusion. By tracking the temporal evolution of PageRank, we identify key influencers, primarily self-identified far-right supporters disseminating hate against liberals and woke culture. Ironically, embracing free speech principles appears to have enabled hate speech against the very concept of freedom of expression and free speech itself. Our findings underscore the delicate balance platforms must strike between open expression and robust moderation to curb the proliferation of hate online.", }

We investigate the impact of free speech and the relaxation of moderation on online social media platforms using Elon Musk{'}s takeover of Twitter as a case study. By curating a dataset of over 10 million tweets, our study employs a novel framework combining content and network analysis. Our findings reveal a significant increase in the distribution of certain forms of hate content, particularly targeting the LGBTQ+ community and liberals. Network analysis reveals the formation of cohesive hate communities facilitated by influential bridge users, with substantial growth in interactions hinting at increased hate production and diffusion. By tracking the temporal evolution of PageRank, we identify key influencers, primarily self-identified far-right supporters disseminating hate against liberals and woke culture. Ironically, embracing free speech principles appears to have enabled hate speech against the very concept of freedom of expression and free speech itself. Our findings underscore the delicate balance platforms must strike between open expression and robust moderation to curb the proliferation of hate online.

X-posing Free Speech: Examining the Impact of Moderation Relaxation on Online Social Networks

woah-1.15

Poster

2404.11465

https://aclanthology.org/2024.woah-1.16.bib

https://aclanthology.org/2024.woah-1.16/

@inproceedings{arora-etal-2024-uli, title = "The Uli Dataset: An Exercise in Experience Led Annotation of o{GBV}", author = "Arora, Arnav and Jinadoss, Maha and Arora, Cheshta and George, Denny and {Brindaalakshmi} and Khan, Haseena and Rawat, Kirti and {Div} and {Ritash} and Mathur, Seema", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.16", doi = "10.18653/v1/2024.woah-1.16", pages = "212--222", abstract = "Online gender-based violence has grown concomitantly with the adoption of the internet and social media. Its effects are worse in the Global majority where many users use social media in languages other than English. The scale and volume of conversations on the internet have necessitated the need for automated detection of hate speech and, more specifically, gendered abuse. There is, however, a lack of language-specific and contextual data to build such automated tools. In this paper, we present a dataset on gendered abuse in three languages- Hindi, Tamil and Indian English. The dataset comprises of tweets annotated along three questions pertaining to the experience of gender abuse, by experts who identify as women or a member of the LGBTQIA+ community in South Asia. Through this dataset, we demonstrate a participatory approach to creating datasets that drive AI systems.", }

Online gender-based violence has grown concomitantly with the adoption of the internet and social media. Its effects are worse in the Global majority where many users use social media in languages other than English. The scale and volume of conversations on the internet have necessitated the need for automated detection of hate speech and, more specifically, gendered abuse. There is, however, a lack of language-specific and contextual data to build such automated tools. In this paper, we present a dataset on gendered abuse in three languages- Hindi, Tamil and Indian English. The dataset comprises of tweets annotated along three questions pertaining to the experience of gender abuse, by experts who identify as women or a member of the LGBTQIA+ community in South Asia. Through this dataset, we demonstrate a participatory approach to creating datasets that drive AI systems.

The Uli Dataset: An Exercise in Experience Led Annotation of oGBV

woah-1.16

Poster

2311.09086

https://huggingface.co/papers/2311.09086

https://aclanthology.org/2024.woah-1.17.bib

https://aclanthology.org/2024.woah-1.17/

@inproceedings{nirmal-etal-2024-towards, title = "Towards Interpretable Hate Speech Detection using Large Language Model-extracted Rationales", author = "Nirmal, Ayushi and Bhattacharjee, Amrita and Sheth, Paras and Liu, Huan", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.17", doi = "10.18653/v1/2024.woah-1.17", pages = "223--233", abstract = "Although social media platforms are a prominent arena for users to engage in interpersonal discussions and express opinions, the facade and anonymity offered by social media may allow users to spew hate speech and offensive content. Given the massive scale of such platforms, there arises a need to automatically identify and flag instances of hate speech. Although several hate speech detection methods exist, most of these black-box methods are not interpretable or explainable by design. To address the lack of interpretability, in this paper, we propose to use state-of-the-art Large Language Models (LLMs) to extract features in the form of rationales from the input text, to train a base hate speech classifier, thereby enabling faithful interpretability by design. Our framework effectively combines the textual understanding capabilities of LLMs and the discriminative power of state-of-the-art hate speech classifiers to make these classifiers faithfully interpretable. Our comprehensive evaluation on a variety of social media hate speech datasets demonstrate: (1) the goodness of the LLM-extracted rationales, and (2) the surprising retention of detector performance even after training to ensure interpretability. All code and data will be made available at https://github.com/AmritaBh/shield.", }

Although social media platforms are a prominent arena for users to engage in interpersonal discussions and express opinions, the facade and anonymity offered by social media may allow users to spew hate speech and offensive content. Given the massive scale of such platforms, there arises a need to automatically identify and flag instances of hate speech. Although several hate speech detection methods exist, most of these black-box methods are not interpretable or explainable by design. To address the lack of interpretability, in this paper, we propose to use state-of-the-art Large Language Models (LLMs) to extract features in the form of rationales from the input text, to train a base hate speech classifier, thereby enabling faithful interpretability by design. Our framework effectively combines the textual understanding capabilities of LLMs and the discriminative power of state-of-the-art hate speech classifiers to make these classifiers faithfully interpretable. Our comprehensive evaluation on a variety of social media hate speech datasets demonstrate: (1) the goodness of the LLM-extracted rationales, and (2) the surprising retention of detector performance even after training to ensure interpretability. All code and data will be made available at https://github.com/AmritaBh/shield.

Towards Interpretable Hate Speech Detection using Large Language Model-extracted Rationales

woah-1.17

Poster

2403.12403

https://huggingface.co/papers/2403.12403

https://aclanthology.org/2024.woah-1.18.bib

https://aclanthology.org/2024.woah-1.18/

@inproceedings{brate-etal-2024-bayesian, title = "A {B}ayesian Quantification of Aporophobia and the Aggravating Effect of Low{--}Wealth Contexts on Stigmatization", author = "Brate, Ryan and Van Erp, Marieke and Van Den Bosch, Antal", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.18", doi = "10.18653/v1/2024.woah-1.18", pages = "234--243", abstract = "Aporophobia, a negative social bias against poverty and the poor, has been highlighted asan overlooked phenomenon in toxicity detec-tion in texts. Aporophobia is potentially im-portant both as a standalone form of toxicity,but also given its potential as an aggravatingfactor in the wider stigmatization of groups. Asyet, there has been limited quantification of thisphenomenon. In this paper, we first quantifythe extent of aporophobia, as observable in Red-dit data: contrasting estimates of stigmatisingtopic propensity between low{--}wealth contextsand high{--}wealth contexts via Bayesian estima-tion. Next, we consider aporophobia as a causalfactor in the prejudicial association of groupswith stigmatising topics, by introducing peoplegroup as a variable, specifically Black people.This group is selected given its history of be-ing the subject of toxicity. We evaluate theaggravating effect on the observed n{--}grams in-dicative of stigmatised topics observed in com-ments which refer to Black people, due to thepresence of low{--}wealth contexts. We performthis evaluation via a Structural Causal Mod-elling approach, performing interventions onsimulations via Bayesian models, for three hy-pothesised causal mechanisms.", }

Aporophobia, a negative social bias against poverty and the poor, has been highlighted asan overlooked phenomenon in toxicity detec-tion in texts. Aporophobia is potentially im-portant both as a standalone form of toxicity,but also given its potential as an aggravatingfactor in the wider stigmatization of groups. Asyet, there has been limited quantification of thisphenomenon. In this paper, we first quantifythe extent of aporophobia, as observable in Red-dit data: contrasting estimates of stigmatisingtopic propensity between low{--}wealth contextsand high{--}wealth contexts via Bayesian estima-tion. Next, we consider aporophobia as a causalfactor in the prejudicial association of groupswith stigmatising topics, by introducing peoplegroup as a variable, specifically Black people.This group is selected given its history of be-ing the subject of toxicity. We evaluate theaggravating effect on the observed n{--}grams in-dicative of stigmatised topics observed in com-ments which refer to Black people, due to thepresence of low{--}wealth contexts. We performthis evaluation via a Structural Causal Mod-elling approach, performing interventions onsimulations via Bayesian models, for three hy-pothesised causal mechanisms.

A Bayesian Quantification of Aporophobia and the Aggravating Effect of Low–Wealth Contexts on Stigmatization

woah-1.18

Poster

https://aclanthology.org/2024.woah-1.19.bib

https://aclanthology.org/2024.woah-1.19/

@inproceedings{dementieva-etal-2024-toxicity, title = "Toxicity Classification in {U}krainian", author = "Dementieva, Daryna and Khylenko, Valeriia and Babakov, Nikolay and Groh, Georg", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.19", doi = "10.18653/v1/2024.woah-1.19", pages = "244--255", abstract = "The task of toxicity detection is still a relevant task, especially in the context of safe and fair LMs development. Nevertheless, labeled binary toxicity classification corpora are not available for all languages, which is understandable given the resource-intensive nature of the annotation process. Ukrainian, in particular, is among the languages lacking such resources. To our knowledge, there has been no existing toxicity classification corpus in Ukrainian. In this study, we aim to fill this gap by investigating cross-lingual knowledge transfer techniques and creating labeled corpora by: (i){\textasciitilde}translating from an English corpus, (ii){\textasciitilde}filtering toxic samples using keywords, and (iii){\textasciitilde}annotating with crowdsourcing. We compare LLMs prompting and other cross-lingual transfer approaches with and without fine-tuning offering insights into the most robust and efficient baselines.", }

The task of toxicity detection is still a relevant task, especially in the context of safe and fair LMs development. Nevertheless, labeled binary toxicity classification corpora are not available for all languages, which is understandable given the resource-intensive nature of the annotation process. Ukrainian, in particular, is among the languages lacking such resources. To our knowledge, there has been no existing toxicity classification corpus in Ukrainian. In this study, we aim to fill this gap by investigating cross-lingual knowledge transfer techniques and creating labeled corpora by: (i){\textasciitilde}translating from an English corpus, (ii){\textasciitilde}filtering toxic samples using keywords, and (iii){\textasciitilde}annotating with crowdsourcing. We compare LLMs prompting and other cross-lingual transfer approaches with and without fine-tuning offering insights into the most robust and efficient baselines.

Toxicity Classification in Ukrainian

woah-1.19

Poster

2404.17841

https://aclanthology.org/2024.woah-1.20.bib

https://aclanthology.org/2024.woah-1.20/

@inproceedings{poudhar-etal-2024-strategy, title = "A Strategy Labelled Dataset of Counterspeech", author = "Poudhar, Aashima and Konstas, Ioannis and Abercrombie, Gavin", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.20", doi = "10.18653/v1/2024.woah-1.20", pages = "256--265", abstract = "Increasing hateful conduct online demands effective counterspeech strategies to mitigate its impact. We introduce a novel dataset annotated with such strategies, aimed at facilitating the generation of targeted responses to hateful language. We labelled 1000 hate speech/counterspeech pairs from an existing dataset with strategies established in the social sciences. We find that a one-shot prompted classification model achieves promising accuracy in classifying the strategies according to the manual labels, demonstrating the potential of generative Large Language Models (LLMs) to distinguish between counterspeech strategies.", }

Increasing hateful conduct online demands effective counterspeech strategies to mitigate its impact. We introduce a novel dataset annotated with such strategies, aimed at facilitating the generation of targeted responses to hateful language. We labelled 1000 hate speech/counterspeech pairs from an existing dataset with strategies established in the social sciences. We find that a one-shot prompted classification model achieves promising accuracy in classifying the strategies according to the manual labels, demonstrating the potential of generative Large Language Models (LLMs) to distinguish between counterspeech strategies.

A Strategy Labelled Dataset of Counterspeech

woah-1.20

Poster

https://aclanthology.org/2024.woah-1.21.bib

https://aclanthology.org/2024.woah-1.21/

@inproceedings{lee-etal-2024-improving, title = "Improving Covert Toxicity Detection by Retrieving and Generating References", author = {Lee, Dong-Ho and Cho, Hyundong and Jin, Woojeong and Moon, Jihyung and Park, Sungjoon and R{\"o}ttger, Paul and Pujara, Jay and Lee, Roy Ka-wei}, editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.21", doi = "10.18653/v1/2024.woah-1.21", pages = "266--274", abstract = "Models for detecting toxic content play an important role in keeping people safe online. There has been much progress in detecting overt toxicity. Covert toxicity, however, remains a challenge because its detection requires an understanding of implicit meaning and subtle connotations. In this paper, we explore the potential of leveraging references, such as external knowledge and textual interpretations, to enhance the detection of covert toxicity. We run experiments on two covert toxicity datasets with two types of references: 1) information retrieved from a search API, and 2) interpretations generated by large language models. We find that both types of references improve detection, with the latter being more useful than the former. We also find that generating interpretations grounded on properties of covert toxicity, such as humor and irony, lead to the largest improvements", }

Models for detecting toxic content play an important role in keeping people safe online. There has been much progress in detecting overt toxicity. Covert toxicity, however, remains a challenge because its detection requires an understanding of implicit meaning and subtle connotations. In this paper, we explore the potential of leveraging references, such as external knowledge and textual interpretations, to enhance the detection of covert toxicity. We run experiments on two covert toxicity datasets with two types of references: 1) information retrieved from a search API, and 2) interpretations generated by large language models. We find that both types of references improve detection, with the latter being more useful than the former. We also find that generating interpretations grounded on properties of covert toxicity, such as humor and irony, lead to the largest improvements

Improving Covert Toxicity Detection by Retrieving and Generating References

woah-1.21

Poster

https://aclanthology.org/2024.woah-1.22.bib

https://aclanthology.org/2024.woah-1.22/

@inproceedings{cercas-curry-etal-2024-subjective, title = "Subjective Isms? On the Danger of Conflating Hate and Offence in Abusive Language Detection", author = "Cercas Curry, Amanda and Abercrombie, Gavin and Talat, Zeerak", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.22", doi = "10.18653/v1/2024.woah-1.22", pages = "275--282", abstract = "Natural language processing research has begun to embrace the notion of annotator subjectivity, motivated by variations in labelling. This approach understands each annotator{'}s view as valid, which can be highly suitable for tasks that embed subjectivity, e.g., sentiment analysis. However, this construction may be inappropriate for tasks such as hate speech detection, as it affords equal validity to all positions on e.g., sexism or racism. We argue that the conflation of hate and offence can invalidate findings on hate speech, and call for future work to be situated in theory, disentangling hate from its orthogonal concept, offence.", }

Natural language processing research has begun to embrace the notion of annotator subjectivity, motivated by variations in labelling. This approach understands each annotator{'}s view as valid, which can be highly suitable for tasks that embed subjectivity, e.g., sentiment analysis. However, this construction may be inappropriate for tasks such as hate speech detection, as it affords equal validity to all positions on e.g., sexism or racism. We argue that the conflation of hate and offence can invalidate findings on hate speech, and call for future work to be situated in theory, disentangling hate from its orthogonal concept, offence.

Subjective Isms? On the Danger of Conflating Hate and Offence in Abusive Language Detection

woah-1.22

Poster

https://aclanthology.org/2024.woah-1.23.bib

https://aclanthology.org/2024.woah-1.23/

@inproceedings{tonneau-etal-2024-languages, title = "From Languages to Geographies: Towards Evaluating Cultural Bias in Hate Speech Datasets", author = {Tonneau, Manuel and Liu, Diyi and Fraiberger, Samuel and Schroeder, Ralph and Hale, Scott and R{\"o}ttger, Paul}, editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.23", doi = "10.18653/v1/2024.woah-1.23", pages = "283--311", abstract = "Perceptions of hate can vary greatly across cultural contexts. Hate speech (HS) datasets, however, have traditionally been developed by language. This hides potential cultural biases, as one language may be spoken in different countries home to different cultures. In this work, we evaluate cultural bias in HS datasets by leveraging two interrelated cultural proxies: language and geography. We conduct a systematic survey of HS datasets in eight languages and confirm past findings on their English-language bias, but also show that this bias has been steadily decreasing in the past few years. For three geographically-widespread languages{---}English, Arabic and Spanish{---}we then leverage geographical metadata from tweets to approximate geo-cultural contexts by pairing language and country information. We find that HS datasets for these languages exhibit a strong geo-cultural bias, largely overrepresenting a handful of countries (e.g., US and UK for English) relative to their prominence in both the broader social media population and the general population speaking these languages. Based on these findings, we formulate recommendations for the creation of future HS datasets.", }

Perceptions of hate can vary greatly across cultural contexts. Hate speech (HS) datasets, however, have traditionally been developed by language. This hides potential cultural biases, as one language may be spoken in different countries home to different cultures. In this work, we evaluate cultural bias in HS datasets by leveraging two interrelated cultural proxies: language and geography. We conduct a systematic survey of HS datasets in eight languages and confirm past findings on their English-language bias, but also show that this bias has been steadily decreasing in the past few years. For three geographically-widespread languages{---}English, Arabic and Spanish{---}we then leverage geographical metadata from tweets to approximate geo-cultural contexts by pairing language and country information. We find that HS datasets for these languages exhibit a strong geo-cultural bias, largely overrepresenting a handful of countries (e.g., US and UK for English) relative to their prominence in both the broader social media population and the general population speaking these languages. Based on these findings, we formulate recommendations for the creation of future HS datasets.

From Languages to Geographies: Towards Evaluating Cultural Bias in Hate Speech Datasets

woah-1.23

Poster

2404.17874

https://huggingface.co/papers/2404.17874

https://aclanthology.org/2024.woah-1.24.bib

https://aclanthology.org/2024.woah-1.24/

@inproceedings{ng-etal-2024-sghatecheck, title = "{SGH}ate{C}heck: Functional Tests for Detecting Hate Speech in Low-Resource Languages of {S}ingapore", author = "Ng, Ri Chi and Prakash, Nirmalendu and Hee, Ming Shan and Choo, Kenny Tsu Wei and Lee, Roy Ka-wei", editor = {Chung, Yi-Ling and Talat, Zeerak and Nozza, Debora and Plaza-del-Arco, Flor Miriam and R{\"o}ttger, Paul and Mostafazadeh Davani, Aida and Calabrese, Agostina}, booktitle = "Proceedings of the 8th Workshop on Online Abuse and Harms (WOAH 2024)", month = jun, year = "2024", address = "Mexico City, Mexico", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2024.woah-1.24", doi = "10.18653/v1/2024.woah-1.24", pages = "312--327", abstract = "To address the limitations of current hate speech detection models, we introduce SGHateCheck, a novel framework designed for the linguistic and cultural context of Singapore and Southeast Asia. It extends the functional testing approach of HateCheck and MHC, employing large language models for translation and paraphrasing into Singapore{'}s main languages, and refining these with native annotators. SGHateCheck reveals critical flaws in state-of-the-art models, highlighting their inadequacy in sensitive content moderation. This work aims to foster the development of more effective hate speech detection tools for diverse linguistic environments, particularly for Singapore and Southeast Asia contexts.", }

To address the limitations of current hate speech detection models, we introduce SGHateCheck, a novel framework designed for the linguistic and cultural context of Singapore and Southeast Asia. It extends the functional testing approach of HateCheck and MHC, employing large language models for translation and paraphrasing into Singapore{'}s main languages, and refining these with native annotators. SGHateCheck reveals critical flaws in state-of-the-art models, highlighting their inadequacy in sensitive content moderation. This work aims to foster the development of more effective hate speech detection tools for diverse linguistic environments, particularly for Singapore and Southeast Asia contexts.

SGHateCheck: Functional Tests for Detecting Hate Speech in Low-Resource Languages of Singapore

woah-1.24

Poster

2405.01842