Multi-domain Tweet Corpora for Sentiment Analysis: Resource Creation and Evaluation

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5046–5054 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5046 Multi-domain Tweet Corpora for Sentiment Analysis: Resource Creation and Evaluation Mamta1, Asif Ekbal1, Pushpak Bhattacharyya1, Shikha Srivastava2, Alka Kumar 2, Tista Saha2 1 Department of Computer Science and Engineering Indian Institute of Technology Patna, India 2 Centre for Development of Telematics (C-DOT, India) {mamta 1921cs11, asif, pb}@iitp.ac.in, {shikha, alkakm, tista}@cdot.in Abstract Due to the phenomenal growth of online content in recent time, sentiment analysis has attracted attention of the researchers and developers. A number of benchmark annotated corpora are available for domains like movie reviews, product reviews, hotel reviews, etc. The pervasiveness of social media has also lead to a huge amount of content posted by users who are misusing the power of social media to spread false beliefs and to negatively influence others. This type of content is coming from the domains like terrorism, cyber security, technology, social issues, etc. Mining of opinions from these domains is important to create a socially intelligent system to provide security to the public and to maintain the law and order situations. To the best of our knowledge, there is no publicly available tweet corpora for such pervasive domains. Hence, we firstly create a multi-domain tweet sentiment corpora and then establish a deep neural network based baseline framework to address the above mentioned issues. Annotated corpus has Cohen’s Kappa measurement for annotation quality of 0.770, which shows that the data is of acceptable quality. We are able to achieve 84.65% accuracy for sentiment analysis by using an ensemble of Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), and Gated Recurrent Unit (GRU). Keywords: Sentiment, Multi-domain Corpus, Deep Learning 1. Introduction With the tremendous increase in the number of users on so- cial media, a huge amount of content is being generated ev- ery day. Twitter generates on average 6000 tweets per sec- ond, 500 million per day, and around 200 billion tweets a year 1. Mining opinions expressed in these tweets can pro- vide interesting insights for the construction of a socially intelligent system. Human creativity and perversity contin- uously generate a large volume of a variety of texts. But, it is also true that many users misuse power of social media to accentuate the sectarian conflict by spreading false be- liefs. Many negative groups use social media platforms to strategize, to gain supporters via tweets, and to negatively influence others. Global politics has been the new victim of social media, impacting the election result. Domains which are contributing to generate such type of contents include terrorism, cyber security, technology, and many other so- cial issues. Mining of opinions from the above-mentioned domains can help the government and security agencies to monitor the content generated everyday. It can help to mon- itor terrorist groups, domestic threats, and crime activities to provide security to public and to maintain the law and or- der. The very …

Content truncated to 500 words