Enhancing Sentiment Analysis Using Domain-Specific Lexicon: A Case Study on GST

Abstract

Sentiment analysis has become a vital classification task because of the massive content over the Internet. Sentiment analysis using lexicon based approach have been used to successfully analyse the sentiment of the user. However, some general-purpose lexicons like SentiWordNet contains words that are objective but they exhibit some polarity in a particular domain. Hence analysing sentiment becomes difficult and inaccurate as some words being polar in a speific domain are misunderstood as neutral words. This paper, proposes a strategy to construct a domain-specific lexicon from a news corpora. For the domain-specific words that are objective in SentiWordNet we adopt an information retrieval technique to assign them with polarity. We also introduce an algorithm using bigrams to analyse the sentiment for evaluating the effectiveness of the proposed lexicon. The experiment showed that our lexicon is effective than SentiWordNet in the particular domain and it improves the accuracy of the sentiment classification.