Natural language processing and information extraction: qualitative analysis of financial news articles

Quantitative financial data are today largely analyzed by automatic computer programs based on traditional or artificial intelligence techniques. Differently, qualitative data and, in particular, articles from on-line news agencies or from financial newspapers are not yet successfully processed. As a result, financial operators suffer from qualitative data-overload. The paper addresses the issue of the use of natural language processing and, in particular, information extraction, for processing qualitative financial data. The financial information extraction system under development at the University of Durham can identify specific kinds of information within a source article, producing a set of relevant templates which represent the most important information in the article and therefore reducing the operators' qualitative data-overload. The application has been designed in close contact with experts of the financial sector and can be fully customized by the user who can add new templates to the existing ones.