{
"paper_id": "2021",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T06:07:40.389153Z"
},
"title": "L3CubeMahaSent: A Marathi Tweetbased Sentiment Analysis Dataset",
"authors": [
{
"first": "Atharva",
"middle": [],
"last": "Kulkarni",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Pune Institute of Computer Technology",
"location": {
"settlement": "Pune"
}
},
"email": ""
},
{
"first": "Manali",
"middle": [],
"last": "Likhitkar",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Pune Institute of Computer Technology",
"location": {
"settlement": "Pune"
}
},
"email": ""
},
{
"first": "Gayatri",
"middle": [],
"last": "Kshirsagar",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Pune Institute of Computer Technology",
"location": {
"settlement": "Pune"
}
},
"email": ""
},
{
"first": "Raviraj",
"middle": [],
"last": "Joshi",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Pune Institute of Computer Technology",
"location": {
"settlement": "Pune"
}
},
"email": "ravirajoshi@gmail.com"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Sentiment analysis is one of the most fun damental tasks in Natural Language Process ing. Popular languages like English, Arabic, Russian, Mandarin, and also Indian languages such as Hindi, Bengali, Tamil have seen a sig nificant amount of work in this area. How ever, the Marathi language which is the third most popular language in India still lags be hind due to the absence of proper datasets. In this paper, we present the first major publicly available Marathi Sentiment Analysis Dataset L3CubeMahaSent. It is curated using tweets extracted from various Maharashtrian person alities' Twitter accounts. Our dataset consists of \u223c16,000 distinct tweets classified in three broad classes viz. positive, negative, and neu tral. We also present the guidelines using which we annotated the tweets. Finally, we present the statistics of our dataset and base line classification results using CNN, LSTM, ULMFiT, and BERTbased deep learning mod els.",
"pdf_parse": {
"paper_id": "2021",
"_pdf_hash": "",
"abstract": [
{
"text": "Sentiment analysis is one of the most fun damental tasks in Natural Language Process ing. Popular languages like English, Arabic, Russian, Mandarin, and also Indian languages such as Hindi, Bengali, Tamil have seen a sig nificant amount of work in this area. How ever, the Marathi language which is the third most popular language in India still lags be hind due to the absence of proper datasets. In this paper, we present the first major publicly available Marathi Sentiment Analysis Dataset L3CubeMahaSent. It is curated using tweets extracted from various Maharashtrian person alities' Twitter accounts. Our dataset consists of \u223c16,000 distinct tweets classified in three broad classes viz. positive, negative, and neu tral. We also present the guidelines using which we annotated the tweets. Finally, we present the statistics of our dataset and base line classification results using CNN, LSTM, ULMFiT, and BERTbased deep learning mod els.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The use of social media has seen a sharp upward trend in recent years. It plays a big role in forming and shaping the views of people on various issues. From sharing facts and opinions to voicing dis sent and grievances, the platform has gained popu larity amongst many users (Nielsen and Schr\u00f8der, 2014) . Twitter is a significant social media plat form. It has been quite popular in India for the past few years. It has been used by many politicians, journalists, and activists to connect with people di rectly. These kinds of interactions are generally strong on emotions, and can be used for develop ing sentiment analysis systems Paroubek, 2010\u037e Mathew et al., 2019) . Such systems have proven to be important for political analysis as well as identifying and curbing more complex issues such as fake news, harassment, hate speech, and bullying (Schmidt and Wiegand, 2017\u037e Joshi et al., 2021\u037e Wani et al., 2021 . In this work, we consider basic sentiment analysis or polarity identification tasks.",
"cite_spans": [
{
"start": 276,
"end": 304,
"text": "(Nielsen and Schr\u00f8der, 2014)",
"ref_id": "BIBREF22"
},
{
"start": 635,
"end": 671,
"text": "Paroubek, 2010\u037e Mathew et al., 2019)",
"ref_id": null
},
{
"start": 850,
"end": 915,
"text": "(Schmidt and Wiegand, 2017\u037e Joshi et al., 2021\u037e Wani et al., 2021",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Popular languages such as English, Arabic, Rus sian, Mandarin (Rogers et al., 2018\u037e Nabil et al., 2015\u037e Yu et al., 2020 as well as Indian languages such as Hindi, Bengali and Tamil have been ex plored on the sentiment task for a long time (Arora, 2013\u037e Patra et al., 2015\u037e Akhtar et al., 2016\u037e Mukku and Mamidi, 2017\u037e Ravishankar and Raghunathan, 2017 . Many resources such as properly anno tated datasets, SentiWordNets, annotation guide lines have been developed for these languages (Socher et al., 2013\u037e Saif et al., 2013\u037e Hu and Liu, 2004 . Alternatively due to the low resource na ture of many languages, translated versions of the English datasets were used for analysis (Joshi et al., 2019\u037e Refaee and Rieser, 2015\u037e Mohammad et al., 2016 . However, such translated datasets are often noisy due to the limitation of current translation systems for low resource languages.",
"cite_spans": [
{
"start": 62,
"end": 119,
"text": "(Rogers et al., 2018\u037e Nabil et al., 2015\u037e Yu et al., 2020",
"ref_id": null
},
{
"start": 239,
"end": 351,
"text": "(Arora, 2013\u037e Patra et al., 2015\u037e Akhtar et al., 2016\u037e Mukku and Mamidi, 2017\u037e Ravishankar and Raghunathan, 2017",
"ref_id": null
},
{
"start": 485,
"end": 542,
"text": "(Socher et al., 2013\u037e Saif et al., 2013\u037e Hu and Liu, 2004",
"ref_id": null
},
{
"start": 677,
"end": 744,
"text": "(Joshi et al., 2019\u037e Refaee and Rieser, 2015\u037e Mohammad et al., 2016",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Marathi is an Indian language spoken by around 83 million people and ranks as the third most spo ken language in India. But surprisingly, there is no significant work or resource for the task of senti ment analysis in Marathi (Kulkarni et al., 2021) . A sentiment analysis dataset curated by IITBombay is available, but it has a very small size consist ing of only 150 samples (Balamurali et al., 2012) . In this paper, we present L3CubeMahaSent 1 the largest publicly available Marathi Sentiment Anal ysis dataset to date. This dataset is gathered using Twitter. Our work is summarized as follows:",
"cite_spans": [
{
"start": 226,
"end": 249,
"text": "(Kulkarni et al., 2021)",
"ref_id": "BIBREF15"
},
{
"start": 377,
"end": 402,
"text": "(Balamurali et al., 2012)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "1. We present a \u223c16,000 tweets strong Marathi Sentiment Analysis Dataset, manually tagged into three classes viz. positive, negative, and neutral.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "2. We provide a comprehensive annotation pol icy useful for tagging sentences by their sen timent. We also provide statistics for our dataset and a balanced split for experimenta tion.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "3. We present the result of our experiments on this dataset on recent deep learning ap proaches to create a benchmark for future comparisons.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Sentiment analysis is a fundamental task of Nat ural Language Processing (Medhat et al., 2014) . The absence of a proper sentiment analysis dataset for the Marathi language has led to limited re search in this area. In this section, we will review some of the works introducing data resources in In dian and other languages. Balamurali et al. 2012presented an approach for crosslingual sentiment analysis using linked wordnets for Marathi and Hindi languages. For this purpose, they used vari ous blogs and travel editorials as a dataset which consisted of about 75 positive and 75 negative reviews. The WordNet approach showed an im provement of 1415 percent over the approach us ing a bilingual dictionary. The Marathi dataset cre ated in this work is very small and cannot be used to train existing deep learning algorithms. Bhardwaj et al. (2020) presented a hostility detec tion dataset in Hindi. Data was collected from various online platforms like Twitter, Facebook, Whatsapp, etc., and was benchmarked using ma chine learning algorithms namely, support vector machine (SVM), decision tree, random forest, and logistic regression. They also labeled each hostile post as either fake, hateful, offensive, or defama tion.",
"cite_spans": [
{
"start": 73,
"end": 94,
"text": "(Medhat et al., 2014)",
"ref_id": "BIBREF17"
},
{
"start": 828,
"end": 850,
"text": "Bhardwaj et al. (2020)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Patra et al. 2018presented details of a shared task in a competition on sentiment analysis of code mixed data pairs of HindiEnglish and Bengali English. The best performing team used SVM for sentence classification. The sentiment analysis of codemixed EnglishHindi and EnglishMarathi text is also studied in (Ansari and Govilkar, 2018). Nabil et al. (2015) introduced Arabic sentiments tweets dataset consisting of 10,000 tweets classi fied as objective, subjective positive, subjective negative, and subjective mixed. They tried 4class sentiment analysis as well as 3class sentiment analysis on the dataset and found that the former was more challenging. They also concluded that SVM performed well on the dataset for the task. Rogers et al. (2018) presented RuSentiment, a dataset for sentiment analysis in the Russian lan guage. They performed experiments on their dataset using algorithms like logistic regression, linear SVM, and neural networks. The best per formance was observed in the case of neural net works. They also released the fastText embeddings that they have used for experimentation. Ikoro et al. (2018) presented results of analyzing sentiments of UK energy consumers on Twitter. They proposed a method in which they combined functions from two sentiment lexica. The first lexicon was used to extract the sentimentbearing terms and the negative sentiments. The second lex icon was used to classify the rest of the data. This method improved the accuracy compared to the general method of using one lexicon.",
"cite_spans": [
{
"start": 337,
"end": 356,
"text": "Nabil et al. (2015)",
"ref_id": "BIBREF21"
},
{
"start": 729,
"end": 749,
"text": "Rogers et al. (2018)",
"ref_id": "BIBREF28"
},
{
"start": 1104,
"end": 1123,
"text": "Ikoro et al. (2018)",
"ref_id": "BIBREF9"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "For creating our dataset, we first manually created a list of various famous personalities who actively tweet about current affairs. Twitter profiles were shortlisted based on their frequency, relevance of activity, and degree of the sentiment of the tweets. Hence a majority of the tweets are from political personalities' profiles and activists as they express a wide range of emotions and sentiments. We at tempted to improve the diversity of the points of view contained in the dataset.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Collection",
"sec_num": "3.1"
},
{
"text": "All tweets in this dataset are specifically in the Marathi language. All hashtags, mentions, spe cial symbols, and the occasional English words are kept in the tweets in the publicly available version of the dataset. We think it is best to keep the orig inal dataset unhampered for anyone to experiment on it. However, while performing experiments, we have removed the aforementioned tokens from the tweets during data preprocessing. Also, the dataset does not retain any context of the tweets such as the tweeting profile, time of posting, re gion, etc.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Collection",
"sec_num": "3.1"
},
{
"text": "As far as scraping the tweets is concerned, there are multiple python libraries available. Some of them are Tweepy (the official opensource library provided by Twitter) 2 , GetOldTweets3 3 , Twint 4 , and Snscrape 5 . We have used the Twint library to scrape tweets.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Collection",
"sec_num": "3.1"
},
{
"text": "We have manually labeled the entire dataset into three classes: positive, negative, and neutral. These three classes have been denoted by '1','1', and '0' respectively. The dataset was split among the entire team to tag in parallel. In order to main tain consistency while tagging tweets, we devel oped an annotation policy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Annotation",
"sec_num": "3.2"
},
{
"text": "To begin with, we ensured not to take into account the author of the tweet, thereby eliminating any bias towards any author. Tweets are tagged by a general assumption that they are posted by any ran dom person. Positive emotions such as happiness, gratitude, respect, inspiration, support are tagged as positive. Negative emotions such as hate, dis respect, grief, insult, disagreement, the opposition are tagged as negative. Tweets that do not convey a strong feeling, such as simple facts, statistics, or statements are tagged as neutral.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Annotation",
"sec_num": "3.2"
},
{
"text": "Tweets containing sarcasm, irony which clearly de pict a negative sentiment are tagged as negative. Congratulatory and thankyou tweets are tagged as positive. A tweet that criticizes something or some one, or which states a fact stating an adverse event or reaction is termed negative. However, if the crit icism comes as constructive and healthy, mention ing possible solutions, then it is tagged as positive. Finally, tweets containing mixed sentiments are la beled by the more dominant emotion expressed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Annotation",
"sec_num": "3.2"
},
{
"text": "Even though these rules were laid down, there were some tweets that simply were difficult to tag by a single individual and needed to be reviewed. In such cases, we took a vote amongst the team and tagged the tweet according to the majority votes. Tweets for which no consensus could be formed were removed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Annotation",
"sec_num": "3.2"
},
{
"text": "Some examples of tagged tweets are mentioned for more clarity in Table 3 given in the Appendix.",
"cite_spans": [],
"ref_spans": [
{
"start": 65,
"end": 72,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Dataset Annotation",
"sec_num": "3.2"
},
{
"text": "Initially, we annotated a total of 18,378 tweets. But, in order to ensure that the classes are bal Train 12114 4038 4038 4038 Test 2250 750 750 750 Validation 1500 500 500 500 anced, we randomly selected an equal number of tweets for each class. Hence, the final version of L3CubeMahaSent consists of 15,864 tweets. Ta ble 1 shows classwise distribution and the train testvalidation split. The remaining 2,514 anno tated tweets will also be published along with the dataset. It consists of 2,355 positive and 159 nega tive tweets. These extra tweets have not been used for model evaluation. Commonly occurring words in each class can be visualized in the form of word clouds as shown in Figure 2 .",
"cite_spans": [],
"ref_spans": [
{
"start": 99,
"end": 172,
"text": "Train 12114 4038 4038 4038 Test 2250 750 750 750 Validation 1500",
"ref_id": "TABREF0"
},
{
"start": 696,
"end": 704,
"text": "Figure 2",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Dataset Statistics",
"sec_num": "3.3"
},
{
"text": "We performed 2class and 3class sentiment anal ysis on our dataset. For conducting baseline ex periments on our dataset, hashtags, mentions, and special symbols were removed during preprocess ing. We used some of the widely used text clas sification architectures for sentiment classification (Kulkarni et al., 2021\u037e Kowsari et al., 2019\u037e Kim, 2014\u037e Sun et al., 2019 . The text is tokenized as words or subwords and passed to the algorithms mentioned below:",
"cite_spans": [
{
"start": 292,
"end": 365,
"text": "(Kulkarni et al., 2021\u037e Kowsari et al., 2019\u037e Kim, 2014\u037e Sun et al., 2019",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "\u2022 CNN: The initial embedding layer outputs word embeddings of size 300. These embed dings are passed to a Conv1D layer having 300 filters and kernel size 3. A global max pooling is applied to the output sequences to get a sentence representation. This is then passed on to a dense layer having size 100. A final dense layer having size equal to the number of classes is added to give classifi cation results. We have experimented with various types of embedding layers having ran dom initialization (word and subword), origi nal Facebook fastText embeddings (trainable and nontrainable) (Mikolov et al., 2018) , and Indic fastText embeddings (trainable and nontrainable) by IndicNLP (Kakwani et al., 2020 ).",
"cite_spans": [
{
"start": 587,
"end": 609,
"text": "(Mikolov et al., 2018)",
"ref_id": "BIBREF18"
},
{
"start": 683,
"end": 704,
"text": "(Kakwani et al., 2020",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "\u2022 BiLSTM+GlobalMaxPool: This is similar to the CNN network with Conv1D layer re placed by a BiLSTM layer. Inputs are fed to an embedding layer which outputs word em beddings of size 300. These embeddings are given to a bidirectional LSTM layer with cell size 300 and then output is max pooled over time. A dense layer of size 100 and a subse quent dense layer of size equal to the number of classes complete the architecture. Embed dings same as those mentioned in the CNN section are also experimented with.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "\u2022 ULMFiT: ULMFiT is also a LSTM based model (Howard and Ruder, 2018) . It uses transfer learning which allows the model to be finetuned quickly on the target dataset us ing even a small sample set. We use a publicly available ULMFiT model for the Marathi lan guage released by iNLTK and finetune it on our dataset (Arora, 2020 ).",
"cite_spans": [
{
"start": 44,
"end": 68,
"text": "(Howard and Ruder, 2018)",
"ref_id": "BIBREF7"
},
{
"start": 314,
"end": 326,
"text": "(Arora, 2020",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "\u2022 BERT: The BERT is a transformer based model pretrained on a huge text corpora, which can be finetuned for any target dataset (Devlin et al., 2019) . Many publicly available flavours of BERT are available, and we use two specific multilingual models:",
"cite_spans": [
{
"start": 127,
"end": 148,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "-MultilingualBERT (mBERT)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "-IndicBERT by IndicNLP (Kakwani et al., 2020) For both of these models, we used the CLS token for sequence classification.",
"cite_spans": [
{
"start": 23,
"end": 45,
"text": "(Kakwani et al., 2020)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimentations",
"sec_num": "4.1"
},
{
"text": "We experimented with a variety of architectures such as CNN and BiLSTM for text classification on our dataset along with different embeddings.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "4.2"
},
{
"text": "We have used random word and subword initial izations, and also used pretrained word embed dings made public by Facebook and IndicNLP. Both of these pretrained embeddings were used in trainable and static modes. Along with these ar chitectures, pretrained models such as ULMFiT, mBERT, and IndicBERT were also used.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "4.2"
},
{
"text": "The results from our experiments were in line with previous works done in Marathi ",
"cite_spans": [
{
"start": 74,
"end": 81,
"text": "Marathi",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "4.2"
},
{
"text": "In this paper, we have presented L3CubeMahaSent the first major publicly available dataset for Marathi Sentiment Analysis which consists of \u223c16000 distinct tweets. We also describe the an notation policy which we used for manually label ing the entire dataset. We performed 2class and 3class sentiment classification to provide a bench mark for future studies. The deep learning mod els used for sentiment prediction were CNN, Bi LSTM, ULMFiT, mBERT, and IndicBERT. The publicly available Marathi fastText embeddings were used with wordbased models. We report the best accuracy using IndicBERT and CNN with In dic fastText word embeddings. We hope that our dataset will play a crucial role in advancing NLP research for the Marathi language.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "5"
},
{
"text": "https://github.com/l3cubepune/MarathiNLP",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://www.tweepy.org/ 3 https://pypi.org/project/GetOldTweets3/ 4 https://pypi.org/project/twint/ 5 https://github.com/JustAnotherArchivist/snscrape",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "This section lists some sample annotated tweets as shown inTable 3.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "This work was done under the L3Cube Pune men torship program. We would like to express our gratitude towards our mentors at L3Cube for their continuous support and encouragement.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": null
},
{
"text": " Tag ",
"cite_spans": [],
"ref_spans": [
{
"start": 1,
"end": 4,
"text": "Tag",
"ref_id": null
}
],
"eq_spans": [],
"section": "S.No. Tweet English Translation",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "A hybrid deep learning archi tecture for sentiment analysis",
"authors": [
{
"first": "Ayush",
"middle": [],
"last": "Md Shad Akhtar",
"suffix": ""
},
{
"first": "Asif",
"middle": [],
"last": "Kumar",
"suffix": ""
},
{
"first": "Push",
"middle": [],
"last": "Ekbal",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Pak Bhattacharyya",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of COL ING 2016, the 26th International Conference on Com putational Linguistics: Technical Papers",
"volume": "",
"issue": "",
"pages": "482--493",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Md Shad Akhtar, Ayush Kumar, Asif Ekbal, and Push pak Bhattacharyya. 2016. A hybrid deep learning archi tecture for sentiment analysis. In Proceedings of COL ING 2016, the 26th International Conference on Com putational Linguistics: Technical Papers, pages 482- 493.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Sentiment analysis of mixed code for the translit erated hindi and marathi texts",
"authors": [
{
"first": "Ansari",
"middle": [],
"last": "Mohammed Arshad",
"suffix": ""
},
{
"first": "Sharvari",
"middle": [],
"last": "Govilkar",
"suffix": ""
}
],
"year": 2018,
"venue": "International Journal on Natural Language Computing (IJNLC)",
"volume": "7",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mohammed Arshad Ansari and Sharvari Govilkar. 2018. Sentiment analysis of mixed code for the translit erated hindi and marathi texts. International Journal on Natural Language Computing (IJNLC) Vol, 7.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "inltk: Natural language toolkit for indic languages",
"authors": [
{
"first": "Gaurav",
"middle": [],
"last": "Arora",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2009.12534"
]
},
"num": null,
"urls": [],
"raw_text": "Gaurav Arora. 2020. inltk: Natural language toolkit for indic languages. arXiv preprint arXiv:2009.12534.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Sentiment analysis for hindi lan guage",
"authors": [
{
"first": "Piyush",
"middle": [],
"last": "Arora",
"suffix": ""
}
],
"year": 2013,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Piyush Arora. 2013. Sentiment analysis for hindi lan guage. MS by Research in Computer Science.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Crosslingual sentiment analysis for indian languages using linked wordnets",
"authors": [
{
"first": "Aditya",
"middle": [],
"last": "Ar Balamurali",
"suffix": ""
},
{
"first": "Pushpak",
"middle": [],
"last": "Joshi",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Bhat Tacharyya",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceed ings of COLING 2012: Posters",
"volume": "",
"issue": "",
"pages": "73--82",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "AR Balamurali, Aditya Joshi, and Pushpak Bhat tacharyya. 2012. Crosslingual sentiment analysis for indian languages using linked wordnets. In Proceed ings of COLING 2012: Posters, pages 73-82.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Ami tava Das, and Tanmoy Chakraborty. 2020. Hos tility detection dataset in hindi",
"authors": [
{
"first": "Mohit",
"middle": [],
"last": "Bhardwaj",
"suffix": ""
},
{
"first": "Shad",
"middle": [],
"last": "Md",
"suffix": ""
},
{
"first": "Asif",
"middle": [],
"last": "Akhtar",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ekbal",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2011.03588"
]
},
"num": null,
"urls": [],
"raw_text": "Mohit Bhardwaj, Md Shad Akhtar, Asif Ekbal, Ami tava Das, and Tanmoy Chakraborty. 2020. Hos tility detection dataset in hindi. arXiv preprint arXiv:2011.03588.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Bert: Pretraining of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "Jacob",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "Mingwei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Kenton",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computa tional Linguistics: Human Language Technologies",
"volume": "1",
"issue": "",
"pages": "4171--4186",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jacob Devlin, MingWei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pretraining of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computa tional Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171-4186.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Universal language model finetuning for text classification",
"authors": [
{
"first": "Jeremy",
"middle": [],
"last": "Howard",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "Ruder",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 56th Annual Meeting of the Associ ation for Computational Linguistics",
"volume": "1",
"issue": "",
"pages": "328--339",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jeremy Howard and Sebastian Ruder. 2018. Universal language model finetuning for text classification. In Proceedings of the 56th Annual Meeting of the Associ ation for Computational Linguistics (Volume 1: Long Papers), pages 328-339.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Mining and summa rizing customer reviews",
"authors": [
{
"first": "Minqing",
"middle": [],
"last": "Hu",
"suffix": ""
},
{
"first": "Bing",
"middle": [],
"last": "Liu",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining",
"volume": "",
"issue": "",
"pages": "168--177",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Minqing Hu and Bing Liu. 2004. Mining and summa rizing customer reviews. In Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 168-177.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Analyzing sentiments ex pressed on twitter by uk energy company consumers",
"authors": [
{
"first": "Victoria",
"middle": [],
"last": "Ikoro",
"suffix": ""
},
{
"first": "Maria",
"middle": [],
"last": "Sharmina",
"suffix": ""
},
{
"first": "Khaleel",
"middle": [],
"last": "Malik",
"suffix": ""
},
{
"first": "Riza",
"middle": [],
"last": "Batistanavarro",
"suffix": ""
}
],
"year": 2018,
"venue": "2018 Fifth International Conference on Social Net works Analysis, Management and Security (SNAMS)",
"volume": "",
"issue": "",
"pages": "95--98",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Victoria Ikoro, Maria Sharmina, Khaleel Malik, and Riza BatistaNavarro. 2018. Analyzing sentiments ex pressed on twitter by uk energy company consumers. In 2018 Fifth International Conference on Social Net works Analysis, Management and Security (SNAMS), pages 95-98. IEEE.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Deep learning for hindi text classification: A compari son",
"authors": [
{
"first": "Ramchandra",
"middle": [],
"last": "Joshi",
"suffix": ""
},
{
"first": "Purvi",
"middle": [],
"last": "Goel",
"suffix": ""
},
{
"first": "Raviraj",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2019,
"venue": "International Conference on Intelligent Human Computer Interaction",
"volume": "",
"issue": "",
"pages": "94--101",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ramchandra Joshi, Purvi Goel, and Raviraj Joshi. 2019. Deep learning for hindi text classification: A compari son. In International Conference on Intelligent Human Computer Interaction, pages 94-101. Springer.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Evaluation of deep learn ing models for hostility detection in hindi text",
"authors": [
{
"first": "Ramchandra",
"middle": [],
"last": "Joshi",
"suffix": ""
},
{
"first": "Rushabh",
"middle": [],
"last": "Karnavat",
"suffix": ""
},
{
"first": "Kaustubh",
"middle": [],
"last": "Jira Pure",
"suffix": ""
},
{
"first": "Raviraj",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2101.04144"
]
},
"num": null,
"urls": [],
"raw_text": "Ramchandra Joshi, Rushabh Karnavat, Kaustubh Jira pure, and Raviraj Joshi. 2021. Evaluation of deep learn ing models for hostility detection in hindi text. arXiv preprint arXiv:2101.04144.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "2020. inlpsuite: Mono lingual corpora, evaluation benchmarks and pretrained multilingual language models for indian languages",
"authors": [
{
"first": "Divyanshu",
"middle": [],
"last": "Kakwani",
"suffix": ""
},
{
"first": "Anoop",
"middle": [],
"last": "Kunchukuttan",
"suffix": ""
},
{
"first": "Satish",
"middle": [],
"last": "Golla",
"suffix": ""
},
{
"first": "Avik",
"middle": [],
"last": "Nc Gokul",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Bhattacharyya",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Mitesh",
"suffix": ""
},
{
"first": "Pratyush",
"middle": [],
"last": "Khapra",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Kumar",
"suffix": ""
}
],
"year": null,
"venue": "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings",
"volume": "",
"issue": "",
"pages": "4948--4961",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Divyanshu Kakwani, Anoop Kunchukuttan, Satish Golla, NC Gokul, Avik Bhattacharyya, Mitesh M Khapra, and Pratyush Kumar. 2020. inlpsuite: Mono lingual corpora, evaluation benchmarks and pretrained multilingual language models for indian languages. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings, pages 4948-4961.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Convolutional neural networks for sentence classification",
"authors": [
{
"first": "Yoon",
"middle": [],
"last": "Kim",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Lan guage Processing (EMNLP)",
"volume": "",
"issue": "",
"pages": "1746--1751",
"other_ids": {
"DOI": [
"10.3115/v1/D14-1181"
]
},
"num": null,
"urls": [],
"raw_text": "Yoon Kim. 2014. Convolutional neural networks for sentence classification. In Proceedings of the 2014 Conference on Empirical Methods in Natural Lan guage Processing (EMNLP), pages 1746-1751, Doha, Qatar. Association for Computational Linguistics.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Text classification algorithms: A survey",
"authors": [
{
"first": "Kamran",
"middle": [],
"last": "Kowsari",
"suffix": ""
},
{
"first": "Kiana",
"middle": [
"Jafari"
],
"last": "Meimandi",
"suffix": ""
},
{
"first": "Mojtaba",
"middle": [],
"last": "Hei Darysafa",
"suffix": ""
},
{
"first": "Sanjana",
"middle": [],
"last": "Mendu",
"suffix": ""
},
{
"first": "Laura",
"middle": [],
"last": "Barnes",
"suffix": ""
},
{
"first": "Donald",
"middle": [],
"last": "Brown",
"suffix": ""
}
],
"year": 2019,
"venue": "Information",
"volume": "10",
"issue": "4",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kamran Kowsari, Kiana Jafari Meimandi, Mojtaba Hei darysafa, Sanjana Mendu, Laura Barnes, and Donald Brown. 2019. Text classification algorithms: A survey. Information, 10(4):150.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Experimental evaluation of deep learning models for marathi text classification",
"authors": [
{
"first": "Atharva",
"middle": [],
"last": "Kulkarni",
"suffix": ""
},
{
"first": "Meet",
"middle": [],
"last": "Mandhane",
"suffix": ""
},
{
"first": "Manali",
"middle": [],
"last": "Likhitkar",
"suffix": ""
},
{
"first": "Gayatri",
"middle": [],
"last": "Kshirsagar",
"suffix": ""
},
{
"first": "Jayashree",
"middle": [],
"last": "Jagdale",
"suffix": ""
},
{
"first": "Raviraj",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2101.04899"
]
},
"num": null,
"urls": [],
"raw_text": "Atharva Kulkarni, Meet Mandhane, Manali Likhitkar, Gayatri Kshirsagar, Jayashree Jagdale, and Raviraj Joshi. 2021. Experimental evaluation of deep learning models for marathi text classification. arXiv preprint arXiv:2101.04899.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Spread of hate speech in online social media",
"authors": [
{
"first": "Binny",
"middle": [],
"last": "Mathew",
"suffix": ""
},
{
"first": "Ritam",
"middle": [],
"last": "Dutt",
"suffix": ""
},
{
"first": "Pawan",
"middle": [],
"last": "Goyal",
"suffix": ""
},
{
"first": "Ani",
"middle": [],
"last": "Mukherjee",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 10th ACM confer ence on web science",
"volume": "",
"issue": "",
"pages": "173--182",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Binny Mathew, Ritam Dutt, Pawan Goyal, and Ani mesh Mukherjee. 2019. Spread of hate speech in online social media. In Proceedings of the 10th ACM confer ence on web science, pages 173-182.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Sentiment analysis algorithms and applications: A survey",
"authors": [
{
"first": "Walaa",
"middle": [],
"last": "Medhat",
"suffix": ""
},
{
"first": "Ahmed",
"middle": [],
"last": "Hassan",
"suffix": ""
},
{
"first": "Hoda",
"middle": [],
"last": "Korashy",
"suffix": ""
}
],
"year": 2014,
"venue": "Ain Shams engineering journal",
"volume": "5",
"issue": "4",
"pages": "1093--1113",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Walaa Medhat, Ahmed Hassan, and Hoda Korashy. 2014. Sentiment analysis algorithms and applications: A survey. Ain Shams engineering journal, 5(4):1093- 1113.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Ad vances in pretraining distributed word representations",
"authors": [
{
"first": "Tom\u00e1\u0161",
"middle": [],
"last": "Mikolov",
"suffix": ""
},
{
"first": "\u00c9douard",
"middle": [],
"last": "Grave",
"suffix": ""
},
{
"first": "Piotr",
"middle": [],
"last": "Bojanowski",
"suffix": ""
},
{
"first": "Christian",
"middle": [],
"last": "Puhrsch",
"suffix": ""
},
{
"first": "Armand",
"middle": [],
"last": "Joulin",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the Eleventh International Confer ence on Language Resources and Evaluation",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tom\u00e1\u0161 Mikolov, \u00c9douard Grave, Piotr Bojanowski, Christian Puhrsch, and Armand Joulin. 2018. Ad vances in pretraining distributed word representations. In Proceedings of the Eleventh International Confer ence on Language Resources and Evaluation (LREC 2018).",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "How translation alters sen timent",
"authors": [
{
"first": "M",
"middle": [],
"last": "Saif",
"suffix": ""
},
{
"first": "Mohammad",
"middle": [],
"last": "Mohammad",
"suffix": ""
},
{
"first": "Svet",
"middle": [],
"last": "Salameh",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Lana Kiritchenko",
"suffix": ""
}
],
"year": 2016,
"venue": "Journal of Artificial Intelligence Research",
"volume": "55",
"issue": "",
"pages": "95--130",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Saif M Mohammad, Mohammad Salameh, and Svet lana Kiritchenko. 2016. How translation alters sen timent. Journal of Artificial Intelligence Research, 55:95-130.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Actsa: Annotated corpus for telugu sentiment analysis",
"authors": [
{
"first": "Radhika",
"middle": [],
"last": "Sandeep Sricharan Mukku",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mamidi",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the First Workshop on Building Lin guistically Generalizable NLP Systems",
"volume": "",
"issue": "",
"pages": "54--58",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sandeep Sricharan Mukku and Radhika Mamidi. 2017. Actsa: Annotated corpus for telugu sentiment analysis. In Proceedings of the First Workshop on Building Lin guistically Generalizable NLP Systems, pages 54-58.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Astd: Arabic sentiment tweets dataset",
"authors": [
{
"first": "Mahmoud",
"middle": [],
"last": "Nabil",
"suffix": ""
},
{
"first": "Mohamed",
"middle": [],
"last": "Aly",
"suffix": ""
},
{
"first": "Amir",
"middle": [],
"last": "Atiya",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 2015 conference on empirical methods in natural language processing",
"volume": "",
"issue": "",
"pages": "2515--2519",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mahmoud Nabil, Mohamed Aly, and Amir Atiya. 2015. Astd: Arabic sentiment tweets dataset. In Proceedings of the 2015 conference on empirical methods in natural language processing, pages 2515-2519.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "The relative importance of social media for ac cessing, finding, and engaging with news: An eight country crossmedia comparison",
"authors": [
{
"first": "Kim",
"middle": [
"Christian"
],
"last": "Rasmus Kleis Nielsen",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Schr\u00f8der",
"suffix": ""
}
],
"year": 2014,
"venue": "Digital journalism",
"volume": "2",
"issue": "4",
"pages": "472--489",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rasmus Kleis Nielsen and Kim Christian Schr\u00f8der. 2014. The relative importance of social media for ac cessing, finding, and engaging with news: An eight country crossmedia comparison. Digital journalism, 2(4):472-489.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Twitter as a corpus for sentiment analysis and opinion mining",
"authors": [
{
"first": "Alexander",
"middle": [],
"last": "Pak",
"suffix": ""
},
{
"first": "Patrick",
"middle": [],
"last": "Paroubek",
"suffix": ""
}
],
"year": 2010,
"venue": "LREc",
"volume": "10",
"issue": "",
"pages": "1320--1326",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alexander Pak and Patrick Paroubek. 2010. Twitter as a corpus for sentiment analysis and opinion mining. In LREc, volume 10, pages 1320-1326.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Sentiment analysis of codemixed indian lan guages: an overview of sail_codemixed shared task@ icon2017",
"authors": [
{
"first": "Dipankar",
"middle": [],
"last": "Braja Gopal Patra",
"suffix": ""
},
{
"first": "Amitava",
"middle": [],
"last": "Das",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Das",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1803.06745"
]
},
"num": null,
"urls": [],
"raw_text": "Braja Gopal Patra, Dipankar Das, and Amitava Das. 2018. Sentiment analysis of codemixed indian lan guages: an overview of sail_codemixed shared task@ icon2017. arXiv preprint arXiv:1803.06745.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Shared task on sentiment analysis in indian languages (sail) tweetsan overview",
"authors": [
{
"first": "Dipankar",
"middle": [],
"last": "Braja Gopal Patra",
"suffix": ""
},
{
"first": "Amitava",
"middle": [],
"last": "Das",
"suffix": ""
},
{
"first": "Ra",
"middle": [],
"last": "Das",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Prasath",
"suffix": ""
}
],
"year": 2015,
"venue": "Inter national Conference on Mining Intelligence and Knowl edge Exploration",
"volume": "",
"issue": "",
"pages": "650--655",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Braja Gopal Patra, Dipankar Das, Amitava Das, and Ra jendra Prasath. 2015. Shared task on sentiment analysis in indian languages (sail) tweetsan overview. In Inter national Conference on Mining Intelligence and Knowl edge Exploration, pages 650-655. Springer.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Corpus based sentiment classification of tamil movie tweets using syntactic patterns",
"authors": [
{
"first": "Nadana",
"middle": [],
"last": "Ravishankar",
"suffix": ""
},
{
"first": "Shriram",
"middle": [],
"last": "Raghunathan",
"suffix": ""
}
],
"year": 2017,
"venue": "IIOAB Journal: A Journal of Multidisciplinary Science and Technology",
"volume": "8",
"issue": "2",
"pages": "172--178",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nadana Ravishankar and Shriram Raghunathan. 2017. Corpus based sentiment classification of tamil movie tweets using syntactic patterns. IIOAB Journal: A Journal of Multidisciplinary Science and Technology, 8(2):172-178.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Benchmark ing machine translated sentiment analysis for arabic tweets",
"authors": [
{
"first": "Eshrag",
"middle": [],
"last": "Refaee",
"suffix": ""
},
{
"first": "Verena",
"middle": [],
"last": "Rieser",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 2015 Conference of the North American Chapter of the Association for Com putational Linguistics: Student Research Workshop",
"volume": "",
"issue": "",
"pages": "71--78",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Eshrag Refaee and Verena Rieser. 2015. Benchmark ing machine translated sentiment analysis for arabic tweets. In Proceedings of the 2015 Conference of the North American Chapter of the Association for Com putational Linguistics: Student Research Workshop, pages 71-78.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Rusentiment: An enriched sentiment analysis dataset for social media in russian",
"authors": [
{
"first": "Anna",
"middle": [],
"last": "Rogers",
"suffix": ""
},
{
"first": "Alexey",
"middle": [],
"last": "Romanov",
"suffix": ""
},
{
"first": "Anna",
"middle": [],
"last": "Rumshisky",
"suffix": ""
},
{
"first": "Svitlana",
"middle": [],
"last": "Volkova",
"suffix": ""
},
{
"first": "Mikhail",
"middle": [],
"last": "Gronas",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Gribov",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 27th international conference on computational lin guistics",
"volume": "",
"issue": "",
"pages": "755--763",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Anna Rogers, Alexey Romanov, Anna Rumshisky, Svitlana Volkova, Mikhail Gronas, and Alex Gribov. 2018. Rusentiment: An enriched sentiment analysis dataset for social media in russian. In Proceedings of the 27th international conference on computational lin guistics, pages 755-763.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Evaluation datasets for twitter sentiment analysis: a survey and a new dataset, the stsgold",
"authors": [
{
"first": "Hassan",
"middle": [],
"last": "Saif",
"suffix": ""
},
{
"first": "Miriam",
"middle": [],
"last": "Fernandez",
"suffix": ""
},
{
"first": "Yulan",
"middle": [],
"last": "He",
"suffix": ""
},
{
"first": "Harith",
"middle": [],
"last": "Alani",
"suffix": ""
}
],
"year": 2013,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hassan Saif, Miriam Fernandez, Yulan He, and Harith Alani. 2013. Evaluation datasets for twitter sentiment analysis: a survey and a new dataset, the stsgold.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "A survey on hate speech detection using natural language pro cessing",
"authors": [
{
"first": "Anna",
"middle": [],
"last": "Schmidt",
"suffix": ""
},
{
"first": "Michael",
"middle": [],
"last": "Wiegand",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the fifth international work shop on natural language processing for social media",
"volume": "",
"issue": "",
"pages": "1--10",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Anna Schmidt and Michael Wiegand. 2017. A survey on hate speech detection using natural language pro cessing. In Proceedings of the fifth international work shop on natural language processing for social media, pages 1-10.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "Recursive deep models for se mantic compositionality over a sentiment treebank",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Socher",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Perelygin",
"suffix": ""
},
{
"first": "Jean",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Jason",
"middle": [],
"last": "Chuang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Christopher",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Manning",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Andrew",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Ng",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Potts",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the 2013 conference on empirical meth ods in natural language processing",
"volume": "",
"issue": "",
"pages": "1631--1642",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D Manning, Andrew Y Ng, and Christopher Potts. 2013. Recursive deep models for se mantic compositionality over a sentiment treebank. In Proceedings of the 2013 conference on empirical meth ods in natural language processing, pages 1631-1642.",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "How to finetune bert for text classification?",
"authors": [
{
"first": "Chi",
"middle": [],
"last": "Sun",
"suffix": ""
},
{
"first": "Xipeng",
"middle": [],
"last": "Qiu",
"suffix": ""
},
{
"first": "Yige",
"middle": [],
"last": "Xu",
"suffix": ""
},
{
"first": "Xuanjing",
"middle": [],
"last": "Huang",
"suffix": ""
}
],
"year": 2019,
"venue": "China National Conference on Chinese Computational Linguistics",
"volume": "",
"issue": "",
"pages": "194--206",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chi Sun, Xipeng Qiu, Yige Xu, and Xuanjing Huang. 2019. How to finetune bert for text classification? In China National Conference on Chinese Computational Linguistics, pages 194-206. Springer.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "Evaluating deep learn ing approaches for covid19 fake news detection",
"authors": [
{
"first": "Apurva",
"middle": [],
"last": "Wani",
"suffix": ""
},
{
"first": "Isha",
"middle": [],
"last": "Joshi",
"suffix": ""
},
{
"first": "Snehal",
"middle": [],
"last": "Khandve",
"suffix": ""
},
{
"first": "Vedangi",
"middle": [],
"last": "Wagh",
"suffix": ""
},
{
"first": "Raviraj",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2101.04012"
]
},
"num": null,
"urls": [],
"raw_text": "Apurva Wani, Isha Joshi, Snehal Khandve, Vedangi Wagh, and Raviraj Joshi. 2021. Evaluating deep learn ing approaches for covid19 fake news detection. arXiv preprint arXiv:2101.04012.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "Chsims: A chinese multimodal sentiment analy sis dataset with finegrained annotation of modality",
"authors": [
{
"first": "Wenmeng",
"middle": [],
"last": "Yu",
"suffix": ""
},
{
"first": "Hua",
"middle": [],
"last": "Xu",
"suffix": ""
},
{
"first": "Fanyang",
"middle": [],
"last": "Meng",
"suffix": ""
},
{
"first": "Yilin",
"middle": [],
"last": "Zhu",
"suffix": ""
},
{
"first": "Yixiao",
"middle": [],
"last": "Ma",
"suffix": ""
},
{
"first": "Jiele",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Jiyun",
"middle": [],
"last": "Zou",
"suffix": ""
},
{
"first": "Kaicheng",
"middle": [],
"last": "Yang",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Associa tion for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "3718--3727",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wenmeng Yu, Hua Xu, Fanyang Meng, Yilin Zhu, Yixiao Ma, Jiele Wu, Jiyun Zou, and Kaicheng Yang. 2020. Chsims: A chinese multimodal sentiment analy sis dataset with finegrained annotation of modality. In Proceedings of the 58th Annual Meeting of the Associa tion for Computational Linguistics, pages 3718-3727.",
"links": null
}
},
"ref_entries": {
"FIGREF1": {
"uris": null,
"type_str": "figure",
"text": "Mean length of records per class (in words)",
"num": null
},
"FIGREF2": {
"uris": null,
"type_str": "figure",
"text": "Positive, Neutral and Negative wordclouds",
"num": null
},
"FIGREF3": {
"uris": null,
"type_str": "figure",
"text": "Confusion Matrix for IndicBERT 3 class classification cation experiments, slightly outperforming the In dicBERT. The IndicBERT was the best perform ing model for the more difficult 3class classifica tion experiments. The respective confusion matri ces are shown inFigure 4andFigure 3.",
"num": null
},
"FIGREF4": {
"uris": null,
"type_str": "figure",
"text": "Confusion Matrix for CNN with Indic fast Text (trainable) 2 class classification",
"num": null
},
"TABREF0": {
"html": null,
"text": "Dataset Statistics.",
"type_str": "table",
"content": "",
"num": null
},
"TABREF2": {
"html": null,
"text": "Classification accuracies over different architectures. The 2class accuracy corresponds to positive and negative class only.",
"type_str": "table",
"content": "",
"num": null
}
}
}
}