Designers of text filtering systems can benefit from research in text retrieval, user modeling and a number of other fields. Text filtering is, however, a unique information seeking process that is distinguished by a focus on satisfying relatively stable interests in documents containing text. This report has reviewed progress in the field with particular emphasis on the selection component of the filtering process. Other useful perspectives are offered by Jiang , Mock , Stevens , and Wyle .
Text filtering systems must develop representations of both documents and user interests, they must be endowed with some way of comparing documents with interests, and they must possess some way of using the results of those comparisons to assist the user with document selection. Text retrieval research has produced a number of content-based representations that use the frequency with which terms appear in documents, and social filtering research has produced a complementary set of features based on shared annotations from other users. When combined with implicit or explicit feedback from the user about the documents they have examined, those representations provide a basis for construction of profiles which represent the user's interests. Both text retrieval and machine learning offer techniques for comparing document representations with profiles, and this is an area of active research. Document visualization is another dynamic research area, but ranked output presently offers a simple way of synergistically exploiting the strengths of human and machine to facilitate the filtering process.
The text filtering techniques described in this report offer a range of solutions that can help users achieve their information seeking goals. With technology presently in hand, designers can produce effective and efficient systems that will be useful in a number of applications. Furthermore, the present research on applications of user modeling, implicit feedback, shared annotations and document visualization to text filtering suggests that text filtering technology will have even greater impact in the future. As the quantity of online information continues to increase, text filtering will provide an increasingly important technique for bringing together producers and consumers of information.