OPTIMAL DOCUMENT-INDEXING VOCABULARY FOR MEDLINE

The indexing vocabulary is an important determinant of success in text retrieval. Researchers have compared the effectiveness of indexing us ing free-text and controlled vocabularies in a variety of text context s. A number of studies have investigated the relative merits of free-t ext, MeSH and UMLS Metathesaurus indexing vocabularies for MEDLINE doc ument indexing. Most of these studies conclude that controlled vocabul aries offer no advantages in retrieval performance over free-text. Thi s paper offers a detailed analysis of prior results and their underlyi ng experimental designs. The analysis indicates that there are a numbe r of open questions relevant to the overall debate on indexing vocabul aries for MEDLINE. This paper also offers results from a new experimen t assessing eight different retrieval strategies. These strategies inv olve document indexing via free-text, MeSH and several alternative com binations of the two vocabularies. The results indicate that MeSH does have an important role in text retrieval. Copyright (C) 1996 Elsevier Science Ltd