Estudo comparativo de algumas linguagens de indexação: eficacia e tempo de pesquisa

Abstract

Report a comparative test on information retrieval by KWIC, Uniterm, Subject Headings, Dewey Decimal Classification, Universal Decimal Classification and Classification of Librarianship and Information Sciences. 100 Brazilian documents on library and information science were selected for the test and every document was indexed according to the 6 systems covered by the evaluation. Eight librarians prepared 10 questions that could be answered by one or more documents from the test collection. Sixty students from the Faculdade de Biblioteconomia e Documentação of the Universidade Santa Úrsula, divided randomly into 6 groups of 10, searched the 6 indexes to find documents that answered the 10 questions. Each researcher was instructed to record time spent in search the entries searched, documents selected for examinations, and documents considered to answer the questions posed. To these data was added the number of documents assgned of each term consulted by the searchers. From the data collected were determined the performance figures for each system. Analysis of the data showed that alphabetical index systems (KWIC, Uniterm and Subject Headings) performed better than the Classified indexing languages. Among the latter, the faceted classification had the best performance. Considering degree of success in answwring the test questions, the systems were ranked as followers: Subject Headings (96%), Uniterm (94%), KWIC (85%), Classification of Librarianship and Information Sciences (84%), Universal Decimal Classification (75%) and Dewey Decimal Classification (73%). Analysis of failures showed that most were due to human failures in indexing and searching rather than to any intrinsic defects in the languages themselves. Searchers preferences favoured Subject Headings. Final results showed that the Subject Headings were ranked first according to 5 of the 6 variables studied, and were ranked second on the other variable.