Financial Hot Topic Detection Based On Sequential Relations

In recent years, Internet has been playing a more and more important role inpeople’s daily life. Every day the complicated news events in various forms werereleased on the Internet. Compared with the traditional printed media, such asnewspapers and magazines, people who live in fast-paced life have becomeaccustomed to choose the Internet as the main source of kinds of information insuch a information explosion period.Every day there is so much information provided by every portal websites,news blogs and micro-blogs that we are often overwhelmed and feel fatigued andconfused. However, the fact is that most people pay more attention to the news ofa very field which they are interested in, for example, sports, science andtechnology, finance and so on. Now the information on the Internet is mostlypresented by the traditional form as a news webpage. The news was listed on thewebpage and ordered by the happen time. But what if you want to pay continuousattention to a series of events of a special topic? Sorry, generally speaking youhave no choice but to search the old events happened several days ago by the aidof search engines. Therefore, it’s very significant to classify and gather therelated events of the same topic and then provide them to user by t he order oftime.In consideration of the above facts, the sequential relationship of news andevents of each listed company in financial field was researched in this paper andthe aim is to establish a automatic analysis system of the sequential relations hipof listed company news and events, which can be used to classify historicalevents and track the new events accurately.The main research contents consist of the following aspects, acquisition andclassification of the stock news, extraction of stock news topic, topic keywordsextraction, discovery and tracking of new topic. First of all, the system needsnews acquisition, news preprocessing and classification. This part mainly dealswith all the stock news in A and B stock market of Shanghai and Shenzh en StockExchange. The second task is extracting hot events from the classified news. Itwill be accomplished with the method of text mining. The third part is topickeywords extraction. The aim is to give every topic a recapitulative descriptionafter the analysis of the whole topic. At the last is discovery and tracking of new topic. Gather newly happened events and confirm whether it’s a continuous partof an old topic or not.Based on the stock news texts collected from some influential financialwebsites, an online system aimed to detect and track financial field news topicwas established. The news of each listed companies between Jan1stand Jun10thof2012was used as test corpus to experiment. Finally, the experimental resultsshow that the proposed algorithm model meets the requirements of onlineapplication system.