Dirichlet Process Mixtures Model Based on Variational Inference for Chinese Person Name Disambiguation

Person name ambiguity in Web search results is a very common phenomenon. Although many methods have been proposed for solving the problem of person name ambiguity, their accuracy still must be enhanced in the complex and heterogeneous webpages. We introduce a variational inference algorithm for the Dirichlet process mixtures model (DPMM) for text clustering to disambiguate person name. Experiments on web data from different search engines indicate that our approach consistently outperforms other clustering methods such as K-means clustering and agglomerative hierarchical clustering.