Comparative Analysis and Enhancement of Sentiment Intensity Based Tools

Abstract

Sentiment intensity of a text indicates the strength of its association with positive or negative sentiment, which is measured in terms of continuous numerical value. Predicting sentiment intensities can achieve more fine-grained sentiment analysis as compared to polarity classification. This research paper is focused on sentiment intensity detection and comprises of two main parts; to find and address loopholes through rule-based approach by making a comparison of existing sentiment intensity measuring tools and a technique to enhance sentiment intensity-based classification by a lexicon-based approach to incorporate Word Sense Disambiguation (WSD). There is a strong need to conduct a thorough comparison of sentiment intensity-based methods to understand the potential limitations, advantages, and disadvantages of popular methods. This article aims at filling this gap by presenting a benchmark comparison of five popular sentiment intensity-based analysis methods. The second part is focused on constructing a human-annotated lexicon of adjectives paired with nouns triggering ambiguity, mapped with intensity as well as polarity. This Lexicon is further integrated with Enhanced Vader to measure performance; the result has outperformed with classification f-score 72%, much better than sentiment detection without lexicon.