Grey sentiment analysis using SentiWordNet

Abstract

Sentiment analysis is one of the most important topics in the Natural Language Processing field, aiming to determine whether a text expresses a positive, negative or neutral perception. In most sentiment analysis applications, a central role is played by the sentiment lexicons, which are lexical resources that include lists of tokens, together with the associated polarity score for each token or term. However, such approaches do not take into consideration the fact that a term might have distinct and sometimes even opposite sentiment polarities in different contexts. The present paper uses the grey system theory in order to associate terms with the most likely intervals of polarity, in order to enable a more accurate sentiment understanding, through grey sentiment analysis, even in limited information contexts, such as social media analysis.