Clipping and analyzing news using machine learning techniques

Abstract

Generating press clippings for companies manually requires a considerable amount of resources.W e describe a system that monitors online newspapers and discussion boards automatically.The system extracts, classifies and analyzes messages and generates press clippings automatically, taking the specific needs of client companies into account. Key components of the system are a spider, an information extraction engine, a text classifier based on the Support Vector Machine that categorizes messages by subject, and a second classifier that analyzes which emotional state the author of a newsgroup posting was likely to be in. By analyzing large amount of messages, the system can summarize the main issues that are being reported on for given business sectors, and can summarize the emotional attitude of customers and shareholders towards companies. © Springer-Verlag Berlin Heidelberg 2001.