When searching for textual information over a large collection, it is often not practical to issue a single query which will produce the desired results. We take a two-step approach to solving this problem:

• First, the searcher takes AIM.

• Then, the searcher FIREs off a search.

Our approach is to use the AIMing portion of the search to limit the resources which need to be examined. By describing the general requirements of the search, much of the potential search space can be eliminated. The space can be described in several different dimensions, such as subject area, date, and level-of-complexity.

The FIRE part of the search is used to actually retrieve the relevant documents. By means of a conversational interface, the searcher is able to refine her search with the aid of a user-interface agent. This agent observes feedback from the searcher which indicates the relevance of the retrieved documents and then suggests new search terms which will improve the query.