Research On Named Entity Recognition Of Traditional Chinese Medicine Based On Deep Learning

The text of traditional Chinese medicine covers a large amount of information with medical value.Accurate and rapid extraction of important data information in the text can provide data support for clinical diagnosis and treatment and patients’ self diagnosis,and then lay a foundation for standardizing diagnosis and treatment of traditional Chinese medicine and promoting the improvement of medical standards.As an entity information extraction technology,named entity recognition can realize the structural transformation of a large number of unstructured data stored in text.It plays an important role in the extraction of entity data information of target object and the construction of knowledge map.but,named entity recognition for the field of traditional Chinese medicine,the lack of labeled data and the insufficient use of text data features are the important reasons for the low accuracy of named entity recognition.Therefore,we uses method of based on the deep learning to study the named entity recognition of traditional Chinese medicine.The main research contents are as follows:(1)A data enhancement algorithm based on improved generation countermeasure network is proposed.The traditional data enhancement algorithm takes SeqGAN as the core,while SeqGAN with LSTM as the generator has the problems of limiting the parallelism of the algorithm and slow training speed.Therefore,this paper replaces the LSTM model with the Transformer model to abandon the original loop structure,allow the algorithm to be more parallel and enhance the relationship between the data,so as to improve the quality of the generated data.Secondly,the Lattice LSTM model in the traditional data enhanced named entity recognition method lacks the ability to effective use vocabulary information and capture the characteristics of Chinese characters.Therefore,this paper replaces the Lattice LSTM model with MECT model,and improves the recognition accuracy of named entities by combining a variety of features.This paper uses the data set of traditional Chinese medicine instructions.After many experiments,it is found that the algorithm can further improve the accuracy of traditional Chinese medicine named entity recognition.(2)A SFM named entity recognition framework of traditional Chinese medicine based on improved Transformer is proposed.SFM,namely SeqGAN*-FastBERTMECT,extracts the corresponding feature vector from the generated data obtained by the improved SeqGAN together with the original real data through the FastBERT pre training model,then obtains the vocabulary information features and Chinese character features of the serialized data by the MECT model,and finally completes the sequence decoding and annotation through CRF to identify the corresponding entities.In this paper,the SeqGAN*,FastBERT and MECT all have Transformer model,In order to further optimize the SFM named entity recognition framework of traditional Chinese medicine,this paper introduces the residual attention mechanism into transformer to improve the overall performance of the model.After many experiments,the results show that this method can effectively improve the accuracy rate of named entity recognition of traditional Chinese medicine.