Design of sentiment analysis system based on multilingual social network data

Abstract

The usage of social sites is increasing among a huge variety of users like activists, spammers, trolls and commentators, etc. and the advancement of the World Wide Web (WWW) has enabled the users to post data in the structure of their choice (video, audio, text , etc.) and in the languages of their liking (English, Hindi, Spanish, French, German , etc.). With such flexibility, more users have started using social networking sites extensively. The users belong to different social and economic strata of the society and their academic credentials also vary to a large range. Due to the variation in types of users, the topics of interest on social networking sites vary from being dating, economy, environment, sports, academics, science, film, television, wildlife and politics. Language switching is also one of the aspects of such social media comments probably to address wider audience and to have a more impactful outcome. The opinions of different users provide feedback or information related to a particular field which is further used for Sentiment Analysis (SA). The main objective of the research is on mining tweets from Twitter for Hindi, English and Hinglish languages and performing the SA of the targeted events. The mining of sentiments in Hindi has its specific issues and challenges. Morphologically, Hindi is very rich and it is, order wise, very much a free language as compared to other languages such as English which decisively adds to complexity when user-generated content is handled. Limitation of resources for Hindi language invites new challenges which range from collection and generation of datasets. The objective of the research is to apply text preprocessing techniques on the collected dataset of Hindi, English and Hinglish languages and discuss their merit and demerits as well. The research also targets at building resources- annotated corpora and subjective lexicon for Hindi and Hinglish language. The corpus is annotated on the foundation of polarity of words into three categories i.e. negative, neut