A Novel Knowledge Network Framework For Financial News Navigation

The development of Internet technology has accelerated the speed of information dissemination and exchange, and the network-news reports featured with huge amount of complex and redundant data have become a new way of pushing information. Among them, financial news reports mainly focus on financial information covering financial markets, industry economics and industrial policy. Currently, to the public, reading financial news reports on the Internet is a major way to get financial information. In this context, many financial news retrieval platforms emerge at a historic moment, including regular search engines(e.g., news.baidu and news.google) and financial domain vertical websites(e.g., finance.sina and ce). In these platforms, users can search the relevant news links by keywords or click the corresponding label in the navigation menu to obtain financial information they need.Nevertheless, for users, especially for those engaged or interested in finance, such ways to obtain financial information cannot effectively meet their need that they can get quickly the relevant information on the same news page and then clarify the ins and outs of the current information to assist them to make right decisions in economic activities after browsing a financial news report. Although some of platforms add a module called "similar news" or "further reading" on the news pages, those news reports containing duplicate information and lacking diversity are based-content pushed basically and thus donot meet users’needs. So this thesis built models for financial news reports and explored the restructuring of financial information. The major work and contributions are as follows:1. Digged out the implied semantic information by learning financial news reports’contents and titles, and then proposed a method to extract financial knowledge based on Improved_LDA and multi-class classification.2. Based on the financial knowledge, proposed a novel knowledge network framework for financial news navigation to help users locate rapidly the related information and then clarify the ins and outs of the current information after reading a financial news report.3. Implemented the framework focusing on real estate and embedded it into financial news pages. The user study showed most of users accept and enjoy the new financial news presentation which can better meet their needs. In addition, Perplexity results demonstrated Improved LDA can achieve better convergence and generalization performance.