Fine-Grained Sentiment Analysis Based on Sentiment Disambiguation

Abstract

In this paper research on the problem of dynamic polarity change in review analysis. Firstly, Apriori algorithm is used to expand the sentiment ambiguous words based on context, and construct the sentiment ambiguous lexicon, namely triples of (sentiment object, sentiment word, sentiment polarity). Then make use of the condition random field model (CRFs) extracted emotional elements from comments, to fine-grained sentiment orientation analysis based on the sentiment ambiguous lexicon. Experimental results over product corpus in mobile-phone and computer domains show that the feasibility of the proposed method, and helps improve the accuracy of sentiment analysis.