The Study Of The Sentiment Semantic Representation Based On Frames

With the development of e-commerce, online shopping is used and accepted by more and more people, whereby the number of comments about the commodities also demonstrate this exponential explosion. These kinds of comments, in a certain extent, reflect the subjective opinions of buyers for certain products. It will impact the decisions of the purchasers or businesses. Since the comments are filled with various sentiment terms, it becomes very tedious and time-consuming to read or analyze these comments and feedback. This paper’s study is based on the resource description and construction method of the sentiment semantics on frames, it includes the resource description of the sentiment semantic and the description on vocabulary level. In addition, it also includes sentiment semantic annotation on the text level. Specific details are as follows:Firstly, this paper introduces the basic theory about FrameNet, explains the related concepts such as:the semantic knowledge of the FrameNet, sentence annotation and Lexical Units. It was proposed that using the FrameNet to represent the sentiment semantics in the text on this basis, then design a sentiment semantic model on frames. The model is divided into three layers:text layer, semantic layer and ontology layer. It’s made the description for views holders, thematic element and attribute of sentiment more clearly to express the information of sentiment semantic.Then, we mark the online comment corpus of the mobile phone products and network games based on the model. On the one hand, build the sentiment knowledge bases of the mobile phone and network games comments, including frames, frame elements, frame relation and lexical units etc. The lexical units have polarity and sentiment intensity property. After that, we compare the sentiment knowledge bases in two fields. On the other hand, study the sentiment semantic annotation method on text, including define the scope of the annotation and semantic representation system, etc.Finally, based on the sentiment knowledge bases on frames and the model of sentiment semantic representation, we do an actual semantic annotation work in goods reviews. Then do the test with coverage and annotator agreement to test the effectiveness of knowledge bases and semantic model.This paper proposes a fine-grained sentiment semantic representation model which makes the frame for a representation unit, can express the information of sentiment semantic in the text effectively. The knowledge bases of Frames and semantic annotation corpus in this research can be used as a data resources and technology in the field of the sentiment analysis of natural language, opinion mining and business intelligence.