A signature file method for Korean text retrieval

Many content-based retrieval methods for text have been proposed. Among them, signature file methods are useful when inverted files are not available. However, traditional word-oriented signature file methods cannot retrieve all the relevant text items in the variable-spacing environment such as Korean text. The authors describe a spacing-tolerant Korean signature extraction method using 2-syllable patterns. With this method, all the relevant text items are retrieved. The advantages of this method are presented by mathematical analysis and experimental performance comparison with a word-oriented method