Research On Chinese Named Entity Recognition In Medical Field

With the rapid development of Internet technology,more and more online medical consultation sites have been trusted and relied on by patients.Patients can tell their doctors about their symptoms through online consultation and seek professional answers.Using information extraction and knowledge mapping technology,key entities,entity attributes and relationships between entities in the online medical consultation text can be extracted and stored,providing a basis for the online medical intelligent question answering system and further improving the patient's online medical experience.Among them,named entity recognition technology is the foundation and key technology in information extraction.Therefore,it is of great practical significance to study how to improve the effect of named entity recognition of online medical consultation texts.At present,the research on Chinese named entities recognition in the medical field is still in its infancy.After investigation and analysis,this article finds that the following problems can be improved:(1)Lack of high-quality public data sets for named entity recognition.(2)There is a certain room for improvement in recognition effect.(3)The research and application of language models such as BERT are not deep enough.(4)There are few ways to integrate multiple named entity recognition technologies.In response to the above problems,this article mainly carried out the following work:(1)In view of the current status of the lack of the open-source named entity recognition data set in the medical field,this article uses the medical consultation text in the online consultation website obtained by the crawler to construct a high-quality labeled data set.(2)Analysis of the effect of the BERT model in the task of named entity recognition in the medical field and the effect of BERT fine-tuning and feature-based methods,which lay the foundation for the following research.(3)Innovatively propose the BERT?Lattice LSTM model and apply it to the Chinese named entity recognition task.The BERT?Lattice LSTM model uses the BERT language model as a feature extraction module,and the Lattice LSTM model as the named entity recognition module,and finally adjusts the output result through the CRF layer.The experimental results show that the BERT?Lattice LSTM model can fully combine the advantages of the BERT language model for obtaining character-level latent semantic information,and the Lattice LSTM model for obtaining word-level information,which greatly improves the effectiveness of the Chinese named entity recognition task.(4)In view of the professionalism and domain of the named entity recognition task in the medical field,this article draws on the idea of multiple recalls in the recommendation system,innovatively designs multiple named entity recall pathways,and uses the Light GBM model for fusion.This method can achieve higher recognition accuracy on the online medical consultation text data set constructed in this artile.Compared with the single-model BERT?Lattice LSTM with the best performance on this data set,the recognition accuracy has been significantly improved.In summary,the method proposed in this article can greatly improve the effect of the Chinese named entity recognition task in the medical field,and finally provide a profound guiding significance for the Chinese named entity recognition technology for the medical field.