Expert Systems for Document Retrieval : Problems in Capturing Synonym Relations from the Experts '

A key problem in designing information retrieval systems is making the information contained in the system available quickly and easily, with a minimum of training or preparation required by the user. New lisers may be experts in their fields and have no problems with the mechanics of using the computer; yet they may still have diflicu1ty in retrieving information because they are not familiar with the particular concept structures used in the information system, such as basic terms and synonym relations. In sorne cases, the requestor may fail ta find information simply because he or she refers to a concept different from the ones included in the system. Instead of requiring the user to learu the terminology and structure of the retrieval system, it seems preferable to make the system capable of matching the terIns in the user's query with those already in the bibliographical database. Techniques derived from artilicial intelligence and knowledge-based systems are DOW heing used ta design user-friendly natural-language analysis programs for information retrieval systems in an attempt ta solve these problems. Software packages such as LIFER and CLOUT notify the researcher whenever the system does not recognize a word or phrase included in a query. They prompt the person to deline the new expression in terms of keywords already in the system. When a satisfactory delinition is reached, the new word or phrase is added to the system vocabulary, so that it can be freely used in the future. The researcher