Sentence Level Fine-grained Sentiment Analysis Based On Syntax Analysis

In recent years, public opinion analysis as a hot research topic is the concern of many scholars.As one of the important questions: text sentiment analysis, which has become a focus of research by many scholars to tap. With the enthusiasm of people involved in the Internet, especially the rapid development of microblogging、letters and other new medias widespread popularity, the majority of Internet users turn from passive recipients of information into a information generator,and with natural language analysis technique continuous innovation and development of micro-blog text sentiment analysis of text, such subjectivity also been widely used.The traditional method is mainly based on emotion or sentence level, complete the appropriate sentiment analysis. currently in public opinion monitoring, stock analysts analyze part of the application fields, and reflects some practical value. but with the development of the Internet, users express colloquial, the text is not standardized increasingly clear that traditional methods have not qualified affectively computing task now, therefore, this paper presents a fine-grained syntactic analysis of affective computing method to solve this problem,based on the adoption of new methods and ideas, will accuracy of affective computing a step further.In this paper, affective computing tasks based on fine-grained syntactic analysis conducted in-depth and meticulous research, first to build a massive fine-grained emotion dictionary contains 23,128 words and the emotional value of the corresponding emotion; while the statement modification, reverse structure, constructed modification emotion dictionary, as well as construct contains 591 emoticons microblogging expression dictionary. Combined How Net, using PMI algorithm based forest synonyms given word semantic computing polar angle, then dependency parsing sentences, combined with semantic dependencies between words, designed 24 kinds of emotion phrases matching rules,and finally got through fine-grained affective computing final emotional outcome. By comparing the experimental data with the sliding window method and SOAD method, this algorithm whose recall and correct rate are more than 90%, better than the other two algorithms. Finally, design the fine-grained computing prototype system based on dependency syntax.This paper focuses on the following parts:1, Building a foundational emotional dictionary, after several rounds of proofreading errors, manual annotation of 23,128 emotional words and corresponding strength of the emotional scale according to the polarity, in order to deal with the sentence reversed, modified structure, construct the corresponding modification of the thesaurus, and gives the microblogging expression dictionary, and finally for the emotional word which can not be recognized by dictionary, combined How Net knowledge base, by obtaining PMI algorithm based on emotion word synonyms polar forests and add to the emotional dictionary.2, Emotional phrase extraction. By dependency parsing, find the syntactic structure of sentence, analyze dependencies between words, discussion and analysis the extent of modification, denial modification, dynamic modification, reverse structure, emotional phrase structure, which summed up 24 kinds of emotional phrase extraction template, and gives a detailed description of each template extracted.3, Sentence-level fine-grained affective computing.By setting the emotional phrase template, template matching sentences,combined with a massive dictionary resource,calculation rules are given different templates, complete emotional phrases sentiment analysis, and finally take into account the emotional impact of a specific sentence,analysis set different weights, resulting in a sentence of fine-grained calculations.4, Constructed of fine-grained affective computing prototype system. System includes text preprocessing, emotional phrase extraction, affective computing, interface display,four modules, the system provides a third-party call interface, fine-grained emotion text microblogging meet computing needs.