Identification of Complex Named-Entities in Chinese Queries Using WWW

In spite of the heated discussions of traditional named-entities, complex named-entities like movie names and book names are seldom referred in literature. This work focuses on the recognition of such named-entities, especially those with little context, which is common in real-world queries. Therefore, we suggest using dynamic Web data as the basis of recognition. To solve the problem of response time of Web-based algorithms, we propose a novel preprocessing method that combines rough segmentation and caching techniques. Evaluations show that the methods have not only achieved fairly good performance, but also lowered system response significantly.