Multiple Approximate String Matching

We present a very simple and eecient algorithm for on-line multiple approximate string matching. It uses a previously known counting-based lter 9] that searches for a single pattern by quickly discarding uninteresting parts of the text. Our multi-pattern algorithm is based on the simulation of many parallel lters using bits of the computer word. Our average complexity to search r patterns of length m is O(rn log m= log n), being n is the text size. We can search patterns of diierent length, each one with a diierent number of errors. We show experimentally that our algorithm is competitive with the fastest known algorithms, being the fastest for a wide range of intermediate error ratios. We give the rst average-case analysis of the ltering eeciency of the counting method, applicable also to 9].