Research On Named Entity Recognition For Automotive Field

With the rapid development of the Internet,structured and unstructured text data is growing at a rapid rate.How to extract useful information from massive text data has become the focus of current research.Named entity recognition is one of the sub-tasks of information extraction.It is the underlying technology for extracting useful information from irregular unstructured text.The effect of named entity recognition directly affects many tasks in the field of natural language processing.In modern society,car has become popular,and it is becoming more and more important to extract named entity information from car text data.Therefore,research on named entity recognition for automotive field is of great significance.In the past two decades,artificial intelligence led by deep learning has once again become the trend of the times.In the field of natural language processing,the Recurrent Neural Network(RNN)has gradually become the mainstream method,especially the Gated Recurrent Unit(GRU),which can better capture dependencies with large time step distances in time series.It has been widely used in the field of natural language processing.With the successful application of the Attention mechanism in the field of natural language processing,the academic community has skillfully integrated the Attention mechanism with deep learning,which enables deep learning to have more powerful feature extraction capabilities.This paper first introduces the research background and significance of named entity recognition,introduces the research status of foreign and domestic named entity recognition,and briefly describes the techniques and methods applied in the process of Chinese named entity recognition.In the next three chapters,this paper proposes three models for the problem of named entity recognition in the automotive field,and introduces and analyzes each model in detail,and builds a dataset in the automotive field named entity.Detailed experiments were performed on the set to compare the different parameters and the specific results of the different models.This paper proposes a named entity recognition model based on BLatticeGRU,and introduces the model in detail,and summarizes the training process of the model.The automotive field named entity labeling specification is proposed and the automotive field named entity labeling dataset Automobile-NER is built on this basis.The BLatticeGRU model and the baseline model were compared on the Automobile-NER dataset.Experiments show that the model performs better than the baseline model.On this basis,BLatticeGRU and Transformer are merged to construct the BLatticeGRU-Attention model.The effects of biword and Transformer layers on the performance of the model are verified by experiments.The effectiveness of the BLatticeGRU-Attention model is verified by comparison with the BLatticeGRU model.Finally,this paper combines BLatticeGRU,Transformer and Conditional Random Fields(CRF),proposes a named entity recognition model based on BLatticeGRU-Attention-CRF,and compares this model with other models on the Automobile-NER dataset.And the good effect of this model in the named entity recognition task in the automotive field is verified.The experimental results of 94.05% precision rate,93.92% recall rate and 93.98% F1 value are obtained.