A two-fold rule-based model for aspect extraction

Opinion target extraction or aspect extraction is the most important subtask of the aspect-based sentiment analysis. This task focuses on the identification of the targets of user's opinions or sentiments from online reviews. In the recent years, syntactic patterns-based approaches have performed quite well and produced significant improvement in the aspect extraction task. However, these approaches are heavily dependent on the dependency parsers which produced syntactic relations following the grammatical rules and language constraints. In contemporary, users do not give much importance to these rules and constraints while expressing their opinions about particular product and neither reviewer websites restrict users to do so. This makes syntactic patterns-based approaches vulnerable. Therefore, in this paper, we are proposing a two-fold rules-based model (TF-RBM) which uses rules defined on the basis of sequential patterns mined from customer reviews. The first fold extracts aspects associated with domain independent opinions and the second fold extracts aspects associated with domain dependent opinions. We have also applied frequency- and similarity-based approaches to improve the aspect extraction accuracy of the proposed model. Our experimental evaluation has shown better results as compared with the state-of-the-art and most recent approaches.