A Smoothing Method for a Statistical String Similarity

We often need to measure similarity between objective information when integrating information. We propose an algorithm in this paper for the Bayesian estimation of the parameters of a statistical string similarity model. To do this we need a smoothing technique for the parameter estimation, because a string similarity model usually contains many parameters. We can make a parameter estimation for the statistical similarity model by introducing a Dirichlet prior. The experimental results show that the proposed method is effective enough for approximate matching.