A Filter Algorithm for Approximate String Matching Based on Match-Region Features

Approximate string matching is a basic problem in computer science. It is widely used in various areas. The aim of this study is to improve the speed of approximate string matching. Filter algorithm for approximate string matching is discussed because it is suitable for large-scale text searching. A novel filter algorithm based on match-region features is presented. Firstly, a q-gram index is used to preprocess text. Secondly, both pattern and text are logically divided into blocks of fixed size kq+1, and then new match-region features are extracted from blocks, and the algorithm optimizes the fundamental q-gram filtration criterion by the new features. Finally, the improved method of choosing filtration-region based on QUASARs block addressing is used for filtration. The experimental results demonstrate that the algorithm achieves higher matching speed than that of QUASARs block addressing by way of improving filtration efficiency. In particular, its matching speed is much faster under low error rate. Experiments also reveal the relationship between matching speed and error rate of new algorithm. These results suggest that the algorithm is useful in a system for approximate string matching with low error rate. It is also powerful for long pattern approximate string matching on the condition of fixed edit distance k.