Word Polarity Consistency in Sentiment Analysis Applications Daniela América da Silva, Paulo Marcelo Tasinaffo, Johnny Marques, Luiz Alberto Vieira Dias, Adilson Marques da Cunha

Abstract

 Polarity classification of words is important for applications such as Opinion Mining and Sentiment Analysis. And the opinions expressed on various websites and media (e.g. blogs, newspapers) are an important criterion for the success of a government product or policy. There are numerous works that, given a feeling, analyze the structure of a sentence and/or document to infer its orientation, the holder of an opinion, the sentiment in an opinion, among others. However, several domain-independent sentiment dictionaries were manually or (semi) automatically created, and there are inconsistencies, as the polarity of the words in a sentiment dictionary may not necessarily be consistent (or correct). This paper presents a study on the use of propositional satisfiability problem (SAT) to verify word polarity consistency in sentiment analysis applications.