In view of the shortcomings of the current syntactic relations research in considering mostly subject-predicate relations,limited recognition ability of sentiment words and neglecting implicit feature extraction,an implicit feature extraction method based on syntactic rules and sentiment words was proposed.By means of Chinese sentiment dictionary resources and word vec-tors trained based on external and experimental corpus,a hybrid sentiment dictionary and a product feature dictionary were con-structed respectively.The dictionary and the defined syntactic rules were used to extract explicit features and sentiment words,and implicit features were extracted according to their co-occurrence relations.Experiments were conducted on camera review corpus and compared with existing methods.Results show that the proposed method can effectively extract explicit and implicit features,and has better performance in obtaining comprehensive feature information.
维普
万方数据