Subject Intermediary Prototype for Query Refinement and Browsing

We have been investigating ways in which topic information can be used to improve the results of searches submitted to scientific document collections. The approach illustrated by this prototype is based on the observation that broad searches can return a very large number of responses and that these responses may not even be on the topic that the searcher was trying to access. Given a general query and a set of retrieved documents, we want to provide a way to interactively refine searches using topic information.

We have implemented a prototype system for use in identifying and exploring issues to be resolved. The prototype is interposed as an intermediate stage in a search of the ACM Digital Library. Given a general query and a set of retrieved documents, the prototype currently presents the searcher with a representation of the topical layout of the responses and allows the searcher to select topics of interest or to exclude non-relevant topics.

A fundamental question currently being explored is how to display and interact with the topic layout. Given a large category set (such as the ACM Computing Classification System (CCS) used in the prototype), the topic display could easily become as overwhelming as a long list of returned documents. One possibility currently employed by the prototype is to display only a portion of the classification. Given a hierarchical categorization, we display the portion of the hierarchy that is represented in the responses. Currently, two possible modes of interaction with this display are implemented.

The searcher can prune non-relevant topic areas and exclude documents that have those topics or have only those topics.
The searcher can also affect which subset of the retrieved documents is displayed by selecting relevant topics and specifying combinations of them (e.g. all documents with any of these topics, all of these topics, 3 out of the 5, etc.) to be displayed.

Last modified: Fri Nov 7 12:28:23 1997