Sentiment Polarity Unit Extraction Of Web-based Financial Information Based On Dependency Parsing

In recent years, Web financial review information on the Internet has showed a rapid growth. The Web sentiment analysis of financial information could play an important role on the field of business management decisions and risk management.This thesis first make a more in-depth analysis of word recognition and semantic orientation analysis, evaluation object extraction, the sentiment status of research evaluation unit among Web financial review sentiment. Besides, combined with the financial sector, this thesis puts forward the purpose of this thesis, which is to build the financial sector dictionary extracting emotion polar units.This thesis puts forward the establishment of two dictionaries aimed at the financial sector dictionary.(1) Financial domain sentiment dictionary. This thesis build the basic emotions dictionary on the basis of HIT’s method of thesaurus extension. Considering the dictionaries integrity, this thesis also add the emotional vocabularies in China knowledge resource integrated database. In addition, the financial sector is so special that experts marked "08 Xinhua Finance Dictionary". Ultimately, this thesis has build a finance emotion dictionary, which is made of 324 positive emotion words and 255 negative emotion words.(2) financial domain evaluation object dictionary. Evaluation objects are with singular, and this singular can help determine emotional tendency of evaluation objects in sentences. This thesis establish a dictionary of financial singular objects based on the transition sentence form, and the dictionary is composed of 303 singular evaluation objects.For sentiment polarity unit extraction. This thesis tries to analyze the interdependence sentiment connection between emotional words and evaluation objects from the perspective of dependency grammar. Thus to identify the objects of emotional words. Firstly, from the start of analyzing the nature of words in the flied of finance to ensure the nature of a word, such as adjective, verb and substantive. Secondly, analyze the nature of the words and the probable element in the sentence, combining with dependency grammar understanding. Consequently find the sentiment word to all dependencies between objects route evaluation, which extracts its corresponding evaluation object and evaluation of extended objects. Finally, the design and implementation of financial information Web sentiment polarity unit extraction system, and through Web financial the experimental test mode on an information system to verify the validity of the method.