README.md

---
language:
- en
- es
- ja
- el
widget:
- text: It is great to see athletes promoting awareness for climate change.
datasets:
- cardiffnlp/tweet_topic_multi
- cardiffnlp/tweet_topic_multilingual
license: mit
metrics:
- f1
pipeline_tag: text-classification
---

# tweet-topic-base-multilingual

This model is based on [cardiffnlp/twitter-xlm-roberta-base](https://huggingface.co/cardiffnlp/twitter-xlm-roberta-base) language model trained rained on ~198M multilingual tweets and finetuned for multi-label topic classification in English, Spanish, Japanese, and Greek.

The models is trained using [TweetTopic](https://huggingface.co/datasets/cardiffnlp/tweet_topic_multi) and [X-Topic](https://huggingface.co/datasets/cardiffnlp/tweet_topic_multilingual) datasets (see main [EMNLP 2024 reference paper](https://arxiv.org/abs/2410.03075)).



<b>Labels</b>: 


| <span style="font-weight:normal">0: arts_&_culture</span>           | <span style="font-weight:normal">5: fashion_&_style</span>   | <span style="font-weight:normal">10: learning_&_educational</span>  | <span style="font-weight:normal">15: science_&_technology</span>  |
|-----------------------------|---------------------|----------------------------|--------------------------|
| 1: business_&_entrepreneurs | 6: film_tv_&_video  | 11: music                  | 16: sports               |
| 2: celebrity_&_pop_culture  | 7: fitness_&_health | 12: news_&_social_concern  | 17: travel_&_adventure   |
| 3: diaries_&_daily_life     | 8: food_&_dining    | 13: other_hobbies          | 18: youth_&_student_life |
| 4: family                   | 9: gaming           | 14: relationships          |                          |


## Full classification example

```python
from transformers import AutoModelForSequenceClassification, TFAutoModelForSequenceClassification
from transformers import AutoTokenizer
import numpy as np
from scipy.special import expit

    
MODEL = f"cardiffnlp/tweet-topic-base-multilingual"
tokenizer = AutoTokenizer.from_pretrained(MODEL)

# PT
model = AutoModelForSequenceClassification.from_pretrained(MODEL)
class_mapping = model.config.id2label

text = "It is great to see athletes promoting awareness for climate change."
tokens = tokenizer(text, return_tensors='pt')
output = model(**tokens)

scores = output[0][0].detach().numpy()
scores = expit(scores)
predictions = (scores >= 0.5) * 1


# TF
#tf_model = TFAutoModelForSequenceClassification.from_pretrained(MODEL)
#class_mapping = tf_model.config.id2label
#text = "It is great to see athletes promoting awareness for climate change."
#tokens = tokenizer(text, return_tensors='tf')
#output = tf_model(**tokens)
#scores = output[0][0]
#scores = expit(scores)
#predictions = (scores >= 0.5) * 1

# Map to classes
for i in range(len(predictions)):
  if predictions[i]:
    print(class_mapping[i])

```
Output: 

```
news_&_social_concern
sports
```

## Results on X-Topic
|       | English | Spanish | Japanese | Greek |
|--------------|---------|---------|----------|-------|
| **Macro-F1** | 55.4    | 48.5    | 50.8     | 41.3  |
| **Micro-F1** | 63.5    | 63.3    | 57.8     | 69.8  |






## BibTeX entry and citation info

@inproceedings{antypas-etal-2024-multilingual,
    title = "Multilingual Topic Classification in {X}: Dataset and Analysis",
    author = "Antypas, Dimosthenis  and
      Ushio, Asahi  and
      Barbieri, Francesco  and
      Camacho-Collados, Jose",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.emnlp-main.1123",
    pages = "20136--20152",
    abstract = "In the dynamic realm of social media, diverse topics are discussed daily, transcending linguistic boundaries. However, the complexities of understanding and categorising this content across various languages remain an important challenge with traditional techniques like topic modelling often struggling to accommodate this multilingual diversity. In this paper, we introduce X-Topic, a multilingual dataset featuring content in four distinct languages (English, Spanish, Japanese, and Greek), crafted for the purpose of tweet topic classification. Our dataset includes a wide range of topics, tailored for social media content, making it a valuable resource for scientists and professionals working on cross-linguistic analysis, the development of robust multilingual models, and computational scientists studying online dialogue. Finally, we leverage X-Topic to perform a comprehensive cross-linguistic and multilingual analysis, and compare the capabilities of current general- and domain-specific language models.",
}