An Improved SentiWordNet for Opinion Mining and Sentiment Analysis

Soumya Vaidya; Mohamed Rafi

doi:10.37591/joadms.v1i2.120

Abstract

Opinion Mining and Sentiment Analysis is an emergent research area, spanning over multiple disciplines such as data mining, text mining, etc. Opinion mining is an art of extracting the opinions from the huge set of opinion set or reviews. Sentiment analysis is a type of natural language processing for tracking the mood of the public about a particular product or topic. The existing works of opinion mining used Sentiwordnet as a lexical resource. The major drawback of this existing Sentiwordnet is non-determination of score count, i.e., it doesn’t provide the details of number of positively, negatively and neutrally scored words. This information is necessary because without the knowledge of score count if the further data mining techniques are applied, it may give inaccurate results. To facilitate the opinion mining task, this work focus on design of Improved Sentiwordnet so that it can produce the count of scored words by distinguishing them into positive, negative and neutral words. Experiments are conducted on standard movie review and product review datasets. This work also make use of Stanford POS tagger for tagging the dataset. The counted words can be used to improve the results comparatively better.Keywords: Opinion mining, sentiment analysis, POS tagging, scoring using improved SentiWordNet