A Template Based Hybrid Model for Chinese Personal Name Disambiguation

This paper proposes a template based hybrid model for Chinese Personal Name Disambiguation (CPND). The template makes use of the features of personal role such as discriminating personal name (nickname, stage name), together with the specific context of most frequent words, personal name nearest words named entities, date and time that are effective for this disambiguation task, as well as surrounding context of nominal, verbal and adjectival constituents. The construction of the templates is automatically derived from the articles that maximizes the deviation of different categories of personal names. The extraction algorithm of keyword features based on the distribution of unlabeled data is also proposed in this paper for this challenging task. In addition, an augmented similarity measure for the CPND model has been designed to calculate the similarity between a standard template and an unlabeled text. The final evaluation reveals that the proposed model can achieve the Fmeasure of 75.75% on the test data.