Research And Realization Of Chinese Explanatory Opinion Aggregation For Product Reviews

Chinese product reviews have been widely researched by many scholars, because it contains more emotion and information for opinion. Chinese explanatory opinion is a part of Chinese product review, which contains detailed or deeper explanation for opinions. However, with the huge number of opinion texts and writing is not standardized, resulting in the explanatory opinion expressed diverse, redundancy, which pose a challenge to explanatory opinion summarization. Therefore, Chinese explanatory opinion aggregation emerged. Chinese explanatory opinion aggregation is a task that recognized explanatory opinion which is semantic equality, according to semantic equality to aggregate different explanatory opinion information cluster. Chinese explanatory opinion aggregation is one of the important issues in NLP.This paper is based on Chinese product reviews, and research explanatory opinion aggregation, including product attributes coreference resolution and explanatory opinion paraphrase recognition. Specifically, this paper concerns the following three aspects:(1) Coreference resolution is important to explanatory opinion aggregation, and the task is to find the different ways of expression for one product attribute. Therefore, this thesis builts a system of coreference resolution based on clustering method. In particular, we discuss the selection of clustering characteristics, the improvement of clustering algorithm. The experimental results show that clustering method can effectively solve the coreference resolution.(2) Paraphrase recognition for explanatory opinion is core mission of explanatory opinion aggregation, and is to recognize explanatory opinion for a certain semantic. This paper discuss two methods for explanatory opinion paraphrase recognition, including built binary classifier and similarity clustering. The experimental results show that similarity clustering method can effectively solve Paraphrases recognition.(3) This thesis built a system of Chinese explanatory opinion aggregation. First, achieving product attribute aggregation, and then complete explanatory opinion aggression, fusion product attributes and explanatory opinion, after that adopt the unified form of expression. Finally, we can get concise explanatory opinion information cluster.