HOW DOES DIALOG'S TARGET WORK?

This article reports an attempt to understand how a new non‐Boolean ranked output online search facility works. In December 1993, Data‐Star Dialog released a relevance ranking tool known as TARGET (Dialog 1993b). There is considerable research activity into text retrieval using ranking methods, as instanced by the TREC experiments (Harman 1993), and professional online searchers may wish to know exactly what the ranking algorithm does in order to be able to exploit the facility to best advantage. Though it has not been possible to ‘crack’ the algorithm to the level of calculating its match to the nearest per cent, it is possible to see three or four factors at work in the way TARGET produces items ranked in decreasing order of match with a query. It is emphasised that the analysis presented here is based only on records from one bibliographic database. A more extensive and comparative study of bibliographic and full‐text databases would be needed to provide universal and definitive findings.