Chinese query correction based on the combination of statistics and characteristics

摘要

In the Retrieval System, the query words correction is an important auxiliary in improving query efficiency. In this paper, according to the characteristics of the Chinese language, a candidate set has been generated for each term of the query string. After the cross combination of the candidate sets, the grid of candidates is created. With the characteristic form combining the factor of n-gram statistical model and phonetic similarity, query term hits, and n-gram similarity, the candidates ranking model has been set and generally balanced, then the candidates are sorted to get the optimal correction results. The experiments show that, our query correction model based on the combination of statistics and characteristics has obtained higher correction accuracy and recall rate.