Interactive searching and interface issues in the Okapi best match probabilistic retrieval system

Abstract

We explore some interface design issues raised by the development and evaluation of a highly interactive information retrieval (IR) system based on a probabilistic retrieval model with relevance feedback. The Okapi system uses term frequency weighting functions to display retrieved items in a best match ranked order; it can also find additional items similar to those marked as relevant by the searcher. Its query expansion mechanism has been implemented in several different ways, and subjected to a number of evaluative studies. The object has been to compare the effectiveness of automatic and interactive query expansion in different user interface environments. In automatic query expansion, the system extracts terms from items judged relevant by the searcher and immediately creates a new query to search for more items. In the interactive versions the extracted terms are displayed, and the searcher may intervene in the construction of a modified query for the next phase of the search. These contrasting approaches serve to illustrate the interplay between different principles of human computer interaction and the information retrieval task. The discussion focuses on the nature of interaction in IR and the interrelationship between functional visibility, the user's cognitive loading, and the balance of control between user and system.