Machine learning for knowledge-based document routing: a report on the TREC-2 experiment
Richard M. Tong, Lee A. Appelbaum · Text REtrieval Conference · 1993
This paper contains a description of the experiments performed by Advanced Decision Systems as part of the Second Retrieval Conference (TREC-2). The overall system developed for TREC-2 demonstrates how it can be combined statistically techniques with a commercially available knowledge-based information retrieval system. As in TREC-1, the tool used for the fully automatic construction of routing queries is based on the Classification and Regression Trees (CART) algorithm. However, in, a departure from TREC-1, we have expanded our definition of what constitutes a document featureswithin the CART output can be used as the basis of topic definitions that can be interpreted by the TOPIC retrieval system developed by Veery, Inc