Feature Based opinion mining through ontologies

Abstract—Feature Based opinion mining through ontologies. As e-commerce is becoming more and more popular, the number of customer reviews about a product grows rapidly. So it is difficult for a potential customer to browse through large numbers of reviews for items of interest and make a decision about some products. In addition, for manufacturer, it’s also difficult to keeping track product responses and mining the opinions from customer reviews, this is obviously unfavorable for improving products quality. To support some applications, we proposed a fine-grained approach to extract and summarize a general opinion and its strength based on features from customer reviews of a product. In this paper, we combine the LDA model and the association rules to extract the product features and the corresponding sentiment words of a product, < Final Year Projects > and use cross-validation to prune the extract result. A completely unsupervised approach was used to calculating the strength of the sentiment words, and we rank all the reviews according to the strength, all the customer reviews are expression to the purchaser in the form of summary, the summary we propose is concise and concrete. This paper proposes several novel techniques to perform these work, our experimental results show that these techniques are highly effective.