Speaker: Raksha Sharma, TRDDC

Date/Time/Venue:  26 Feb, 12 noon, SIT001

Title: Exploiting Word Properties for Sentiment Analysis

One of the main tasks that have lured businesses and customers is to extract the polarity of the user-generated content available on the Web in the form of reviews on shopping or opinion sites, posts, blogs or customer feedback. As many users do not explicitly indicate their sentiment polarity, it needs to be predicted from the text which has led to a plethora of work in the field of Sentiment Analysis (SA). More and more sophisticated techniques are built to tackle the problem. The evolution of methods has been on several different dimensions. Some of these are the complexity of the algorithm, the knowledge source used, dealing with unlabeled datasets, etc.

Ideally, the Sentiment Analysis (SA) task is to predict the sentiment orientation of a text (document, paragraph or sentence) by analyzing the polarity (positive or negative) of words present in the text. However, mere identification of a word as a polar word is inadequate for deep (fine-grained) sentiment analysis. In our work, we tackle some of the rare, yet important, problems in sentiment analysis, that is, finding properties of polar words. Identification of properties of polar words allows us to progress in the direction of fine-grained sentiment analysis, which can result beyond zero (negative) - one (positive) polarity. In our work, we have explored four properties, viz., domain dependence for polarity, domain dependence for significance, intensity variation among words having the same semantic property and intensity variation among words having the same sense (synonymous words). We incorporated these properties in various dimensions of sentiment analysis, viz., in-domain sentiment analysis, cross-domain sentiment analysis, star-rating prediction, etc.

Speaker's Bio:
Raksha Shama received the B.Tech. Degree in Information Technology from RGPV Bhopal, in 2006 and the M.Tech degree in Information Technology from IIITM Gwalior, in 2009. She received Ph.D. in Computer Science and Engineering under Prof. Pushpak Bhattacharyya from IIT Bombay, in 2017. She has been working at TATA Research Development and Design Center as Research Scientist since October 2017. Her areas of research include Natural Language Processing and Machine Learning. During her Ph.D., she worked on enrichment of sentiment analysis by exploring various properties of polar words.