Aspect and Sentiment Extraction Based on Information-Theoretic Co-clustering

In this paper, we propose an aspect and sentiment extraction method based on information-theoretic Co-clustering. Unlike the existing feature based sentiment analysis methods, which only process the explicit associations between feature words and sentiment words. Our method considers the implicit associations intra evaluated features, the association intra sentiment words, and the associations inter evaluated features and sentiment words. At first, the co-occurrence relationships of feature words and sentiment words are represented as a feature-sentiment words matrix. And with the feature-sentiment words matrix, the information-theoretic Co-clustering algorithm is used to simultaneously cluster evaluated features and sentiment words. The clustering results of feature words are viewed as different aspects of the evaluated objects, and the clustering results of sentiment words which are associated with different aspects are viewed as aspect specific sentiment words. The experimental results demonstrate that this method can obtain good performance of aspect and sentiment extraction.