Sublinear Approximate String Matching

The present paper deals with the subject of approximate string matching and demonstrates how Chang and Lawler [CL94] conceived a new sublinear time algorithm out of ideas that had previously been known. The problem is to find all locations in a text of length n over a b-letter alphabet where a pattern of length m occurs with up to k differences (substitutions, insertions, deletions). The algorithm will run in O( n m k logbm) time when the text is random and k is bounded by the threshold m/(logbm+O(1)). In particular, when k = o(m/ logbm) the expected running time is o(n).