INFORMATION RETRIEVAL SYSTEM

Information retrieval system supporting academic research

Abstract

We report on a study that investigated the efficacy of four different interactive information retrieval (IIR) systems, each designed to support a specific information-seeking strategy (ISS). These systems were constructed using different combinations of IR techniques (i.e., combinations of different methods of representation, comparison, presentation and navigation), each of which was hypothesized to be well suited to support a specific ISS. We compared the performance of searchers in each such system, designated “experimental,” to an appropriate “baseline” system, which implemented the standard specified query and results list model of current state-of-the-art experimental and operational IR systems. Four within-subjects experiments were conducted for the purpose of this comparison. Results showed that each of the experimental systems was superior to its baseline system in supporting user performance for the specific ISS (that is, the information problem leading to that ISS) for which the system was designed. These results indicate that an IIR system, which intends to support more than one kind of ISS, should be designed within a framework which allows the use and combination of different IR support techniques for different ISSs. Because of the requirements imposed by today's professional life, Information Technology (IT) software becomes increasingly complex. We argue in this paper that it is an advantage to use a framework approach—an advantage not only for operational IR applications but also for research purposes, particularly for the evaluation of indexing and retrieval techniques. The IR framework FIRE is being developed along these lines. The rationale for developing FIRE is discussed, and the structure and some of the basic properties of FIRE are explained.

Information retrieval system supporting academic research

Introduction

It has long been clear that there exist different information-seeking strategies (ISSs) that people engage in during the information seeking process, though traditional information retrieval (IR) systems are designed to support mainly one ISS: specified searching. This suggests that IR systems should be designed to support multiple ISSs. Research has shown that different IR systems can be designed to support different ISSs, such as browsing or specified searching. Belkin (1996) proposed that such different IR systems can be construed as incorporating different combinations of IR support techniques. Some examples of IR support techniques are as follows: for comparison, best match or exact match; for navigation, following links or scanning a list; for representation, manual or automatic indexing or classification; for presentation, clustering or listing (Scott, 2001, 49).

The diversity of ISSs indicates that an IR system supporting one ISS well is unlikely to support the others at the same level. Conversely, an IR system that employs only one combination of techniques to support all ISSs is unlikely to be successful at effectively supporting any one of them. Because it seems to be the case that searchers may engage in different ISSs within the course of a single information-seeking episode, it makes sense to design IR systems that can support a range of ISSs.

We propose that building such an interactive IR (IIR) system might be ...