The Opinion Mining Based on Fuzzy Domain Sentiment Ontology Tree for Product Reviews

In this paper, we present a novel method that integrates domain sentiment knowledge into the analysis approach to deal with feature-level opinion mining By constructing a domain ontology called Fuzzy Domain Sentiment Ontology Tree (FDSOT), we then utilize the prior sentiment knowledge of our ontology to achieve significantly accuracy in sentiment classification. Particularly, the FDSOT is the conceptual model which represents the semantic relation between features and sentiment words. The evaluation is based on the Chinese product reviews collected from 360buy.com. The experimental results demonstrate that our approach is able to automatically identify the domain-dependent polarity for a large subset of sentiment expression and effectively improve the performance of opinion mining.