HindiMD: A Multi-domain Corpora for Low-resource Sentiment Analysis

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 7061–7070 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 7061 HindiMD: A Multi-domain Corpora for Low-resource Sentiment Analysis Mamta1, Asif Ekbal1, Pushpak Bhattacharyya2, Tista Saha3, Alka Kumar 3, Shikha Srivastava3 1 Department of Computer Science and Engineering Indian Institute of Technology Patna, India 2 Department of Computer Science and Engineering Indian Institute of Technology Bombay, India 3 Centre for Development of Telematics (C-DOT, India) {mamta_1921cs11, asif}@iitp.ac.in, pb@cse.iitb.ac.in, {shikha, alkakm, tista}@cdot.in Abstract Social media platforms such as Twitter have evolved into a vast information sharing platform, allowing people from a variety of backgrounds and expertise to share their opinions on numerous events such as terrorism, narcotics and many other social issues. People sometimes misuse the power of social media for their agendas, such as illegal trades and negatively influencing others. Because of this, sentiment analysis has won the interest of a lot of researchers to widely analyze public opinion for social media monitoring. Several benchmark datasets for sentiment analysis across a range of domains have been made available, especially for high-resource languages. A few datasets are available for low-resource Indian languages like Hindi, such as movie reviews and product reviews, which do not address the current need for social media monitoring. In this paper, we address the challenges of sentiment analysis in Hindi and socially relevant domains by introducing a balanced corpus annotated with the sentiment classes, viz. positive, negative and neutral. To show the effective usage of the dataset, we build several deep learning based models and establish them as the baselines for further research in this direction. Keywords: Sentiment, Low-resource Language, Indian Language, Multi-domain, Deep Learning, BERT 1. Introduction Social media has become extremely popular among users to express their opinions on various serious events, issues and trending topics. Sentiment analysis is an important field of study that seeks to investigate the polarity of such user-expressed opinions (Pang et al., 2002). It has a wide range and powerful applications in natural language processing (NLP), from customer service to shaping political campaigns (Pang and Lee, 2004; Blitzer et al., 2007; Malo et al., 2013; Bakliwal et al., 2013; Akhtar et al., 2017). However, emerging social media platforms like Twitter and Facebook have become the new channel of information dissemination for many negative groups for recruitment in order to promote terrorist acts and illegal drug trades 1, incite vi- olence and crime by influencing others. The mining of opinions expressed in these domains provides valuable stores of information that can be used by security agen- cies and the government in national counter-terrorism investigations and monitoring crime, drug activities, and domestic threats to ensure public safety. This infor- mation is also helpful in the defense sector for national security strategies, which are vital to national safety and security. Initial research was focused on product and movie reviews (Kumar et al., 2020) and now broad- ened to the other domains (Mamta et al., 2020), in- …

Content truncated to 500 words