Combine Person Name and Person Identity Recognition and Document Clustering for Chinese Person Name Disambiguation

This paper presents the HITSZ_CITYU system in the CIPS-SIGHAN bakeoff 2010 Task 3, Chinese person name disambiguation. This system incorporates person name string recognition, person identity string recognition and an agglomerative hierarchical clustering for grouping the documents to each identical person. Firstly, for the given name index string, three segmentors are applied to segment the sentences having the index string into Chinese words, respectively. Their outputs are compared and analyzed. An unsupervised clustering is applied here to help the personal name recognition. The document set is then divided into subsets according to each recognized person name string. Next, the system identifies/extracts the person identity string from the sentences based on lexicon and heuristic rules. By incorporating the recognized person identity string, person name, organization name and contextual content words as features, an agglomerative hierarchical clustering is applied to group the similar documents in the document subsets to obtain the final person name disambiguation results. Evaluations show that the proposed system, which incorporates extraction and clustering technique, achieves encouraging recall and good overall performance.