Unsupervised product feature extraction for feature-oriented opinion determination

Identifying product features from reviews is the fundamental step as well as a bottleneck in feature-level sentiment analysis. This study proposes a method of unsupervised product feature extraction for feature-oriented opinion determination. The domain-specific features are extracted by measuring the similarity distance of domain vectors. A domain vector is derived based on the association values between a feature and comparative domain corpora. A novel term similarity measure (PMI-TFIDF) is introduced to evaluate the association of candidate features and domain entities. The results show that our approach of feature extraction outperforms other state-of-the-art methods, and the only external resources used are comparative domain corpora. Therefore, it is generic and unsupervised. Compared with traditional pointwise mutual information (PMI), PMI-TFIDF showed better distinction ability. We also propose feature-oriented opinion determination based on feature-opinion pair extraction and feature-oriented opinion lexicon generation. The results demonstrate the effectiveness of our proposed method and indicate that feature-oriented opinion lexicons are superior to general opinion lexicons for feature-oriented opinion determination.