As stated in the abstract, “The relevance feedback process uses information derived from an initially retrieved set of documents to improve subsequent search formulations and retrieval output.” This process has been shown to be effective in a vector processing environment and in probabilistic information retrieval. Extending relevance feedback to a Boolean retrieval environment presents difficulties. In particular, in a Boolean system it is necessary to choose not only the terms to be used in a reformulated query, but also the Boolean operators relating the terms. This problem is being addressed by researchers with some success, as reported in this paper.

The general material in this paper on relevance feedback makes an excellent, brief introduction to this whole area. Two (previously published) approaches to Boolean query relevance feedback (see [1, 2, 3]) are presented, in algorithmic form, with supporting discussion.

The heart of the paper is a description of the evaluation of these two methods. It is interesting to see the care that is taken in the evaluation method to permit a fair comparison of original and feedback searches. The conclusion, as substantiated by experimental results, is that the disjunctive normal form method of Boolean relevance feedback is viable, and that it should be considered to this recommendation.