TOLERATING FUZZINESS IN KEYWORDS BY SIMILARITY SEARCHES
H.-J. Schek · Kybernetes · 1977
One feature of a user‐friendly system is the capability to tolerate fuzziness in the form of names or keywords. The following remarks are a first step towards a model for the intuitive human‐like notion of similarity. This model is characterized by using only the context within single words for a definition of similarity measures. These measures are based on maximal common substrings and abstract syllables. In order to obtain an efficient computation of this formal similarity in large lists, a preselection method is given which uses a simple distance between strings and a precomputed binary relation between character‐pairs and keywords.