Extracting Opinion Features in Sentiment Patterns

Due to the increasing amount of opinions and reviews on the Internet, opinion mining has become a hot topic in data mining, in which extracting opinion features is a key step. Most of existing methods utilize a rule-based mechanism or statistics to extract opinion features, but they ignore the structure characteristics of reviews. The performance has hence not been promising. A new approach of OFESP (Opinion Feature Extraction based on Sentiment Patterns) is proposed in this paper, which takes into account the structure characteristics of reviews for higher values of precision and recall. With a self-constructed database of sentiment patterns, OFESP matches each review sentence to obtain its features, and then filters redundant features regarding relevance of the domain, statistics and semantic similarity. Experimental studies on real-world data demonstrate that as compared with the traditional method based on a window mechanism, OFESP outperforms it on precision, recall and F-score. Meanwhile, in comparison to the approach based on syntactic analysis, OFESP performs better on recall and F-score.