Exploring Word Similarity to Improve Chinese Personal Name Disambiguation

This paper presents an approach to the Chinese Personal Name Disambiguation (PND). The key to clustering is the similarity measure of context, which depends on the features selection and representation and calculation method. First HIT Tongyici Cilin (Extended) is introduced to Chinese PND to enhance the clustering effect. Exploration about more word similarity is also performed to alleviate the data sparseness. In this system, a HAC (Hierarchical Agglomerative Clustering) algorithm is adopted to cluster the mentions referring to a same person with features extracted from documents. The results show that the word similarity information is very helpful to improve the system´s performance.