Identification of Implicit Features using Persian Language Rules and Sentiments Clustering

Typically, when someone wants to buy an online product, he/she reviews the comments and notes written by others about the product. Clearly, it has a profound impact on the person's decision to buy or not to buy the product. Sentiment analysis or opinion mining is one of the hot research topics in computer science. The main purpose of sentiment analysis is to extract opinions of individuals about the characteristics of an entity such as a product. In this research, an unsupervised opinion mining is proposed for Persian products based on implicit feature extraction which is a critical step in sentiment analysis. In most previous studies, statistical information is utilized to create a co-occurrence matrix and determine the implicit features. In this paper, we benefit from syntactic rules and sentiment clustering in conjunction with statistical information to construct an efficient co-occurrence matrix between features and sentiment words. The evaluation results provided on a real-world dataset, extracted from Digikala website, indicates that the proposed method achieves higher recall and precision compared to the previous studies.