This report details the Okapi ENQUIRE Project, which forms part of a series of studies in the development of an experimental text retrieval system.
The project examined users' query formulation and search strategies in a highly interactive GUI environment. One point for investigation was whether they took advantage of the thesaurus and if so, at what point in the search. It also examined how incremental query expansion (adding to the existing query new words and phrases taken from relevant documents) was accepted by users, whether it helped them to find more items, and whether it could be used to redeem a search which was failing.
A configurable user interface was implemented as an experimental tool to run either on an X-Windows colour terminal, or under X-terminal emulation on a PC. All development work was done in a Unix environment. The ENQUIRE system comprises four main components: an interface layer, a query layer, a basic search system managing a document database, and an Oracle RDBMS managing a thesaurus. The query layer and interface layer software were written especially for the project, in C++ and TCL/TK respectively.
The user interface was designed iteratively by observing and interviewing users of the system at various stages during system development. The project culminated with a three-month field trial during which ENQUIRE was installed at two sites and each unsupervised search session was logged extensively. 560 sessions were logged in all, 450 at the main site (the London Business School) and 110 at City University. Analysis of the log data indicates that query reformulation and expanded searches were undertaken in about 25% of sessions. Incremental query expansion provided a well-controlled method of adding new words and phrases to queries. Suggested thesaurus terms (displayed following input of the initial query) were selected by some users, but there were very few instances of explicit thesaurus navigation.