Named Entity Relation Extraction Based on Multiple Features

For the limited availability of established Mongolian named entity dictionary, with the increase of the new terminology and network vocabulary, Mongolian named entity dictionary was not able to update in time. This paper puts forward a method of named entity recognition, which based on a combination of multiple features and CRF. The method of named entity recognition proposed in this paper firstly cluster the entities that never appear in the repository with the method of hierarchical converged clustering, then carry through the normal words recognition and adjust the results of feature base.