Sentiment Computation Of Web Financial Text Based On Semantic Analysis

With the increasingly close links between the network and the financial industry, the Web financial information is also being widely used to provide high quality market forecast. It is different from the financial statements information, called numerical data, which is easy to analyze and calculated. How to further mine and quantify these seemingly useless junk data, and become the magic weapon for predict the financial crisis of the listed companies? This is the hot direction in recent years:emotion detection of the Web financial information. Text sentiment classification has two types: based on machine learning and semantic-based analysis. And machine learning is simply divided texts into two categories or more classes roughly, without taking into account the semantic relationship of the text internal, and the emotional properties of text fine-grained. Semantic-based classification becomes more and more popular.Sentiment analysis for text based on Semantic which is worked on three levels of vocabulary, sentence, document. The emotional tendency value of the entire document is determined by the sentences’ emotional values, which are decided by the emotional words within the sentences. While there are no ready-made emotional corpora in the financial and securities domain, emotional analysis applied in this field is still rare. The main work of this paper are constructing an emotional detection dictionary for financial domain and calculating emotional tendentiousness of financial information.In the aspect of sentiment computation for the Web financial information, the article improves the method of sentiment computation based solely on the morpheme. Firstly, this paper uses LTP of Harbin Institute to analysis the syntactic pattern of sentence. Analysis and experiment for24syntactic analysis from the quantitative and qualitative angle, respectively. And select the select the six kinds of emotional influential syntactic pattern. Secondly, according to the analysis of the syntactic pattern for emotional impact, design of affective computing rules for the6kinds of syntactic pattern. Finally, describes the sentiment calculation model based on the syntactic pattern make use of tree structure. And put forward the principle, define, construct algorithms, calculation rules for the sentiment model.The writer has experimented two ways to analyze the emotional inclinations of Web financial information. One is based on morpheme, the other is combined the morpheme and syntax tree. The first group of experimental results shows that the way of calculating the value of emotional tendency for text which based on improved morpheme, which can make accurate, recall and precise of Positive and Negative have a more greatly improved than Ku’s. The second group of experimental results shows that combined the morpheme and syntax tree has more greatly improved than the simple algorithm of morpheme.