Set-oriented retrieval

摘要

The broad way in which we look at how an IRS functions influences the types of questions we ask about it and the ways we try to improve performance. In the recent past, retrieval methodologies have been based on retrieving documents one at a time. In this paper we are introducing a set-oriented view. We observe that this view is quite consistent with the single-document or sequential methods, and define a precise model to capture the set-oriented approach. We then examine a number of consequences of the model, such as the limitations implied by a finite index vocabulary. Finally, we discuss various ways in which the set orientation can influence our thinking about IR.