Chinese Named Entity Recognition with Inducted Context Patterns

Since whether or not a word is a name is determined mostly by the context of the word, the context pattern induction plays an important role in name entity recognition (NER). We present a NER method based on the context pattern induction. It induces high-precision context patterns in an unsupervised way starting with some entity seeds. Then it uses directly the matched context patterns, instead of extracted entities by inducted patterns, as the features of a CRF-based NER model. The experiments show that the proposed method improves the performance of the high quality named entity recognizer, and achieves higher accuracy and recall rate.