Effective Comment Sentence Filtering For Feature-Based Opinion Mining Improvement

With the rapid development of modern science technology, feature-based opinion mining, a newly developing field of studying, has become a hot research subject across the world. As a pivotal supplementary mean of music information retrieval, the study of automatic text classification shows great significance both in theory and in practice.The vast Chinese market has the world’s largest Internet users and the largest online shopping population. It’s creating a huge database of product reviews. Product reviews research has great significance in study consumer satisfaction, in study consumer spending habits, in get new consumption growth points and so on. In this dissertation, we present a series of studies on the automatic classification of product reviews text by means of three approaches:feature analysis and feature selection and extraction, design of the classifier.All the data in this dissertation are come from TaoBao, JingDong, Dangdang and other e-commerce businesses website. There are5000reviews in our database. They are divided according to punctuation. After that, we have19,500data, which are the major source of experimental data.Feature-based opinion mining aims to provide fine-grained product feature comment from the product reviews. Previous work proposed a lot of statistic-based and model-based approaches. However, the extraction result is not satisfied when they are actually used into an application with large useless data due to the complexity of Chinese. Through error factor analysis, we found some patterns or models have been misused on some sentences, which are no value for our purpose. This paper focuses on improving the POS-pattern match methodology. The core idea of this approach is filtering out the effective comment sentences before feature and sentiment extraction based on neural network training. Three attributes of sentences are selected to learn the classification algorithm. Experiment gives the optimized parameters for the algorithm. We reported the classification performance and also compare the feature extraction performance with filtering process and not. The result on practical data set demonstrates the effect of this approach.The problem of gettingEffective Comment Sentence can be solved by automatic text classification method.Generally, automatic classification of text can be regarded as a two-stage procedure—feature extraction and categorization. We consider2aspects of words and structure, picking out two representative features which were used in classification experiments. According to the result, BP neural network turned out to be the most effective method when coping with the classification of product reviews.Product reviews analysis and processing can be useful to improve services for e-commerce business website. The study of filtering out the effective comment sentences can greatly benefits the further Feature-based opinion mining retrieval and other processing techniques, so engaging in it turns out to be essential and significant.