Hybrid rule-based approach for aspect extraction and categorization from customer reviews

E-commerce business is becoming more and more popular as the number of customers shopping online is increasing every day. Companies ask their customers to review products and services offered by them over their websites. For the big companies, the number of reviews could be in the thousands. So it is almost impossible for any company to read these reviews manually and find out whether customers liked their product or not. Many techniques have been proposed for sentiment classification of reviews. In this paper we are proposing rule-based hybrid approach which exploits sequential patterns and normalized Google distance (NGD) to extract explicit as well as implicit aspects. For grouping synonyms, we are proposing Google similarity distance in conjunction with particle swarm optimization (PSO).