Distributed objects in a large scale text processing system (industrial case study)

We describe the distributed object oriented architecture of TREVI, a system designed to help overcome the information overload problem. TREVI provides a comprehensive framework for the processing and dissemination of documents coming from news streams and other external sources. TREVI is designed to perform a variety of linguistic and text processing tasks, according to the informational needs expressed in user profiles. Many different user profiles may be simultaneously present, and the sheer volume of information that has to be managed by the system poses serious scalability and performance concerns. We show how the use of a distributed, component based architecture allows us to cope with these challenges, and furthermore provides an excellent base for future system extension, allowing for a rapid response to changing user informational needs