Simultaneously accessing large numbers of text documents is an activity that is not well supported by current search engine interfaces. Many solutions have been explored that employ some form of clustering, or document summarization, to facilitate the assessment of retrieved documents. The research outlined in this paper combines the two solutions to provide summaries of document clusters. The idea itself is not new; what is new is that some attempt has been made to exploit the topical structure of documents in the summarization process. However, this novel aspect is not well described, and other researchers may have difficulty replicating this work based on the description given. Two types of summaries are described: multi-document summaries, which emphasize the similarities between documents within a cluster, and single document summaries, which emphasize the difference between a document and other documents in the same cluster.

A user evaluation, based on a standard information retrieval experimental protocol, is presented to assess the effectiveness of the clustering and summarization techniques. The evaluation compares three interfaces: a standard interface that offers lists of retrieved documents; a clustering interface offering a series of document clusters; and a summarization interface that offers clusters, a summary of each cluster, and a summary of each document. In the published version of the paper, the screen dumps are of such poor quality that they are illegible, so it is not possible to evaluate the quality or readability of the summaries created.

The results are not conclusive, displaying no real user preference for one interface over the others. One of the main findings is that familiarity with the search topic correlates with a preference for the standard list interface, whereas user unfamiliarity with the topic correlates with a preference for the more advanced interfaces. It is a pity that more correlations like this, or qualitative results, were not reported, as they may form a useful guide to why users prefer one interface over another.