Analysis of Product Evaluations: An Adaptive Approach Based on Extended Sentiment Dictionaries

In the past decades, the development of Internet has generated enormous information exchanges. The market strategy benefit a lot from analyzing sentiment tendency hidden in short texts. However, it seems hard to analyze the sentiment tendency hidden in short texts in Chinese because of the complexity of this language. In this paper, an approach for summarizing the product evaluations is presented. An adaptive method of sentiment analysis is applied to do sentiment classification which divides the evaluations into positive set, neutral set and negative set. In this method, the general sentiment dictionaries are extended to be special ones for specific topics. Besides, the degree adverbs and multiple negations are taken into consideration. Simulation results show this new method reaches higher accuracy compared with previous method based on general sentiment dictionaries. Then K-means clustering algorithm is employed to cluster the positive set and negative set. For a clear expression, we draw word cloud pictures to show the key words of each clustering. By analyzing the words shown on the pictures, marketing analysts can adjust their market strategies.