Research On Sentiment Analysis Of E-commerce Review Text

The technology of sentiment analysis is a way to identify the emotional tendency of review texts by the way of mining and analyzing the subjective information, such as standpoints, opinions, emotions and so on, computing the intensity of emotional tendency and judging the sentiment classification of review texts. Recently, with the development and extension of electronic commerce, research on sentiment analysis of e-commerce review texts has gradually become a hot topic. Both the buyer and the seller can know clearly about the advantages and disadvantages of the product. And then the seller can improve the products and formulate appropriate sales strategies and the buyer can get necessary products those actually needed. However, only by using the way of calculating the number of negative words and positive words, the traditional sentiment classification methods cannot acquire the true emotion expressed by the review texts. In this paper, to improve the precision rate of sentiment classification, we respectively use semantic method and the method of machine learning to study the technology of sentiment analysis taking consideration of shortage of traditional sentiment classification methods. Specific work is as follows:1. We use the method of sentiment dictionary to extract the characteristic of review texts, and then we construct classification model to compute the emotional tendency. Finishing analyzing the characteristic of review texts, we consider both modifiers and relationship of collocation to construct sentiment dictionary, and then we put forward a specific method of solving the problem of annotation of emotional intensity. Evaluative elements, sentiment words, privative words, adverbs of degree, adversative conjunctions and relationship of collocation are all included in our sentiment dictionary, and then we calculate and judge the emotional tendency of review texts. The experimental results of contrast experiment proved the effectiveness of modifiers and relationship of collocation.2. We use the method of latent semantic index to transform the characteristic of review texts, and then we construct classification model to compute the emotional tendency. In this paper, to solve the problem of ranking the review texts, we use the method of latent semantic index to construct the model of review texts, achieving the purpose of reducing dimension. And then we construct classification model to rank the review texts by using the method of ordinal regression to train the training sample set. The experimental results of contrast experiment proved the effectiveness of our proposed method.