The model uses the knowledge in the knowledge base to solve the current NER task. Preprocessing and model parameter tuning are also investigated to improve the performance. The e®ect of the model was demonstrated by in-domain and crossdomain experiments, achieving promising results. |