Identification of Informative Reviews Enhanced by Dependency Parsing and Sentiment Analysis

The processing and analysis of the customer reviews have been received increasing attention recently. Filtering the noise and useless sentences from the reviews is the first step for this work. In this paper a new informative review identification method is proposed based on dependency parsing and sentiment analysis. These two linguistic concepts can be used to extract effective features from the sentences in the reviews. The framework for identifying informative reviews is set up, which include two stages: feature extraction and classifier learning. The experiment results on the Amazon product reviews show that the performance of the classification has been improved significantly. The method proposed in this paper can be used as a preprocessing in various review analysis and information extraction.