Showing papers on "Thesaurus (information retrieval) published in 1969"

Word-word associations in document retrieval systems

Abstract: The SMART automatic document retrieval system is used to study association procedures for automatic content analysis. The effect of word frequency and other parameters on the association process is investigated through examination of related pairs and through retrieval experiments. Associated pairs of words usually reflect localized word meanings, and true synonyms cannot readily be found from first or second order relationships in our document collections. There is little overlap between word relationships found through associations and those used in thesaurus construction, and the effects of word associations and a thesaurus in retrieval are independent. The use of associations in retrieval experiments improves not only recall, by permitting new matches between requests and documents, but also precision, by reinforcing existing matches. In our experiments, the precision effect is responsible for most of the improvement possible with associations. A properly constructed thesaurus, however, offers better performance than statistical association methods.