Fuzzy Matching as a Retrieval-Enabling Technique for Digital Libraries
http://www.ecst.csuchico.edu/~luk/Papers/AsisMay96Paper.html

Girill, T.R. and Clement H. Luk
CSU Chico
1996

Abstract:

This paper advocates an often-neglected search-support technique, approximate or 'fuzzy' matching of user search terms. When properly deployed, fuzzy matching can significantly enhance the benefits of other, more common approaches to end-user answer retrieval from online reference collections. We compare crude with more sophisticated approximation techniques to explain how astute fuzzy-match software can convert many different near-miss situations (such as those involving faulty prefixes or suffixes, character misplacement, nonstandard word stems, or unanticipated redescription of concepts) into more adequate results. We also suggest practical ways to overcome fuzzy matching's own major drawbacks (namely, problems with search speed, search imprecision, and misinterpretation of search results). The resulting analysis clarifies how to deploy fuzzy matching for maximum effectiveness. We conclude that appropriate fuzzy matching enables more frequent, more flexible search success than do ordinary retrieval-improvement techniques used without it.