期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：17
页码：4211
出版社：Journal of Theoretical and Applied
摘要：Sentiment analysis of product reviews is an attracting and increasing interest in the area of natural language processing and web text mining. Objective is to analyze the effect of ANN based method for opinion classification. In the research that has done so far on sentiment analysis, ANNs have been considered rarely. In this work, the probabilistic neural network (PNN) has been examined in sentiment classification. This work also examines neural network based sentiment classification methods for feature level sentiment classification on various levels of word granularity are used as features. Product reviews collected from the Amazon reviews website are used as dataset for evaluation. Our objective is to classify the product reviews into three classes: positive, negative and neutral. The results are empirically compared with SVM using various quality measures. The superiority of PNN with Principal Component Analysis (PCA) is also shown in terms of training time. PNN is found to perform better and yields higher accuracy in prediction. In general, statistical based approaches do not perform well as that of neural network based approaches. Compared with traditional techniques, the ANN based approach shows the performance improvement in quality measures and in training time. Through the experimental results it will be show that shortening of training time and increasing the classification accuracy can be achieved by hybrid combination of PNN with PCA.
关键词：Opinion Mining; Classification; Principal Component Analysis; Neural Networks; Sentiment Analysis
