Continuous Multi-Source Information Gathering and Classifi–

This paper describes a fully functional prototype of a multi-component system that allows users to retrieve, store and search documents from a variety of publicly available information sources, in a variety of languages, on any subject domain users may be interested in. The system integrates a crawler, which selects and downloads potentially relevant documents, a cleaning tool which removes irrelevant information from the retrieved documents (e.g. advertisements), a document filter, a document classifier, and several other tools which extract meta-information from texts such as titles and keywords. Unlike ad-hoc search engines, the system satisfies long-term information needs of users since it continuously collects documents they may be interested in. The functionality of the system has been evaluated by comparing it with a traditional manual press clipping service, and it proved to give good results.