Abstract: Traditional identification techniques focus on a single domain and lack the mutual collaboration among different domains， which often lead to dumb results. So， a topic domain identification model combining local and global characteristics is proposed.Local identification is performed based on entities’ local characteristics within one domain. On the other hand， these local identification results tend to be consistent with each other based on the global characteristics such as the collaboration and relevance among domains， which can maintain the accuracy of identification effectively. In addition， some improvements are made for the algorithm of topic domain identification， including similarity matrix updating， collaboration quantifying and iteration terminating. The experiments demonstrate the feasibility and effectiveness of the proposed model.

Key words: topic domain identification, mutual collaboration, local characteristics, global characteristics, domain relevance