Evaluation of Spanish thesauri for information retrieval

Abstract

The article describes an experimental test developed in Spanish databases. The test compares the performance between different thesauri and the natural language, in the process of information retrieval. The test was carried out with the combination of the analytic and sampling methods. The searches were made interactively, the records were evaluated, and the search strategies modified based on the parcial results. The references obtained with each kind of language were evaluated on the basis of the precision and recall ratios. The results point out that free text obtained a precision ratio of 63.4 % and a recall of 59.5 %; with the help of the thesauri both ratios improved: the precision reached a 86.8 % and the recall a 61.6 %. The conclusion is that the controlled vocabulary experiences lower levels of noise than the free text, at the same time it can develop high ratios of recall as the controlled one. Also, it is possible to conclude that the controlled vocabulary neutralizes the deficiences of the free text and vice versa, so both are complementary.