Generate domain-specific sentiment lexicon for review sentiment analysis

Abstract

Lexicon-based approaches for review sentiment analysis have attracted significant attention in recent years. Lots of sentiment lexicon generation methods have been proposed. However, the generation of domain-specific lexicon with unlabeled data has not been effectively addressed. In this paper, we propose a new domain-specific sentiment lexicon generation method, mutual information is introduced to assign terms with Part-Of-Speech (POS) tags in the lexicon, the training data are selected from unlabeled corpus according to their sentiment scores which are evaluated by the SentiWordNet (SWN) based sentiment classifier. Then we propose a completed lexicon-based sentiment analysis framework which uses the domain-specific sentiment lexicon generated by the proposed domain-specific sentiment lexicon generation method. The experiment is carried out on publically available datasets. Results show that the proposed lexicon-based sentiment analysis framework using domain-specific lexicons generated by the proposed method gets a good performance.