An interface has been implemented for exploring the structure of document sets returned in response to a query. The interface allows a user to find subsets of documents that are especially relevant to the query through interaction with an interactive dendrogram which displays a hierarchical clustering of the documents. Dynamic lists of document titles are interlocked with the dendrogram to provide detail of the clusters being viewed by the user. For efficiency, the interface has been implemented in several parallel and distributed computation environments and has been applied to retrieval of encyclopedia articles and news stories. KEYWORDS Clustering, information retrieval, interfaces. 1. INTRODUCTION Classical information retrieval systems are based on the idea that each document in a collection can be compared to a user's query, scored as to its similarity, and displayed in similarity-sorted order. However, this linear sorting may obscure structure among the documents. For instan...