The Research Of Chinese Named Entity Recognition

Named Entity Recognition(NER)plays an important role in Natural Language Processing(NLP),it has been widely used in automatic question answering,reading comprehension,knowledge graph and machine translation and so on.With the development of natural language processing technology and research of text data mining,it becomes very important to acquire semantic knowledge in text,named entity recognition is the cornerstone of information application technology such as event or relationship extraction,and plays an indispensable role in the process of text information structuring.Traditional machine learning methods rely on people's domain knowledge and features extracted manually when dealing with this task.In order to obtain better results without hand-crafted features,in this thesis,we propose an NER method based on model ensemble of BiLSTM.Firstly,we apply the BiLSTM-CRF training on the data,and the character-based model Char-NER and the word-based model Word-NER are obtained.Then the score vectors obtained by the two models are operated and concatenated,and we send the concatenated vectors as features to the final classifier for training,using final classifier for model ensemble of Char-NER and Word-NER.The experimental results show that this method achieves good results on the 1998 people's daily and MSRA corpus without hand-crafted features.On the basis of the ensemble model's results,this thesis propose an organization recognition method that incorporates chapter information,by calculating the mutual information in the text to correct the recognition result.The experimental results show that this method has a good performance on organization recognition,after correcting the recognition results,the F-score on People's Daily and MSRA corpus reached 89.72% and 85.15% respectively.