The invention discloses a method and device for building a name corpus recognition module and belongs to the field of information. The method comprises the steps that each modeling Chinese corpus is labeled for at least two different name corpus training modules; when consistency of prediction results of the modeling Chinese corpuses reaches a preset index according to the labeling result, the modeling Chinese corpuses are added into a corpus training list; the corpus features of each modeling Chinese corpus in the corpus training list are extracted; according to the corpus features of the modeling Chinese corpuses, a name corpus recognition model is built. By means of the multiple different name corpus training models, each Chinese corpus is labeled, and the name corpus recognition model is built based on the labeling results. In the process, manual labeling of the user is not needed, the corpus labeling cost is reduced, the labeling results of the name corpus training models are combined when the name corpus training model is built, and the recognition precision of the built name corpus recognition model is improved.