Text-learning and related intelligent agents: a survey

This is the authors' abstract. We don't add key points for this paper.

In surveying current research in the development of text-learning intelligent agents, the author focuses on three key criteria: what representation the particular application uses for documents, how it selects features, and what learning algorithm it uses. She then describes Personal WebWatcher, a content-based intelligent agent that uses text-learning for user-customized Web browsing.