Chinese NER Using CRFs and Logic for the Fourth SIGHAN Bakeoff

We report a high-performance Chinese NER system that incorporates Conditional Random Fields (CRFs) and first-order logic for the fourth SIGHAN Chinese language processing bakeoff (SIGHAN-6). Using current state-of-theart CRFs along with a set of well-engineered features for Chinese NER as the base model, we consider distinct linguistic characteristics in Chinese named entities by introducing various types of domain knowledge into Markov Logic Networks (MLNs), an effective combination of first-order logic and probabilistic graphical models for validation and error correction of entities. Our submitted results achieved consistently high performance, including the first place on the CityU open track and fourth place on the MSRA open track respectively, which show both the attractiveness and effectiveness of our proposed model.