Contextual Sentiment Analysis in Social Media Using High-Coverage Lexicon

摘要: Automatically generated sentiment lexicons offer information for a large number of terms and often at more granular level than manually ones. While such rich has the potential enhancing analysis, it also presents challenge finding best possible strategy to utilising information. In SentiWordNet, negation lexical valence shifters (i.e. intensifier diminisher terms) are associated with scores. Therefore, could either be treated as sentiment-bearing using scores offered by lexicon, or modifiers that influence assigned adjacent terms. this paper, we investigate suitability both these approaches applied classification. Further, explore role non-lexical common social media introduce score aggregation named SmartSA. Evaluation on three datasets show is effective outperform baseline aggregate-and-average approach.