A Context-Dependent Sentiment Analysis of Online Product Reviews based on Dependency Relationships

Consumers often view online consumer product review as a main channel for obtaining product quality information. Existing studies on product review sentiment analysis usually focus on identifying sentiments of individual reviews as a whole, which may not be effective and helpful for consumers when purchase decisions depend on specific features of products. This study proposes a new feature-level sentiment analysis approach for online product reviews. The proposed method uses an extended PageRank algorithm to extract product features and construct expandable context-dependent sentiment lexicons. Moreover, consumers’ sentiment inclinations toward product features expressed in each review can be derived based on term dependency relationships. The empirical evaluation using consumer reviews of two different products shows a higher level of effectiveness of the proposed method for sentiment