Research On Business Information Of News Text

With the continuous development of Internet media, various news media provides a convenient channel for people to learn about the world. News text has its own advantages as one of a large number of media, the news text contains lots of business information, if we can access the business information quickly and effectively, which will undoubtedly provide strong favorable support for the business decision-makers that make plans and the enterprise that control market dynamics. Chinese industrial TAOBAO marketing is working on the research of Business Information, having a lot of information associated with it. This thesis is based on it to further mining, mainly from two aspects which are the news topic sentence extraction and the news elements extraction.This paper starts from the two main characteristics of news text which are the news headlines that reflect the news idea and the important sentence that is placed in the front of other sentences. The importance of the text sentence is measured by the position of the sentence, the overlap ratio and relevancy degree between sentence and title. The overlap ratio takes the number and importance of overlap words into consideration, and relevancy degree via calculating the title and the sentence weight matrix, using the maximum matching algorithm of weighted bipartite graph to get the score of every sentence. Eventually using the way of feature weighted to combine the features and ranking the score to extract the topic sentence. The final experiment shows that the accuracy rate of topic sentence extraction in this paper is 75.9%.In the news element extraction, according to the topic sentences and headlines of news we present a who-driven elements extraction method. Firstly obtaining the news subjects who by Ranking SVM, then creating patterns to get the other elements. According to the experimental results that apply this method to get the elements can reflect the main information of news to some extent.Finally, the system that about business information mining of news text is designed and implemented. The system sets up index for news resources, which can dig out business information which the user is interested in according to user needs, including text, topic sentences and elements of news.