Research And Application Of Biomedical Named Entity Recognition Based On Reinforcement Learning

The amount of biomedical literature is vast and it is of great significance to mine valuable biomedical information.Biomedical named entity recognition is to identify relevant biomedical entity information from the unstructured literature,which is an important prerequisite for biomedical relationship extraction and construction of biomedical knowledge graph.At present,most of the research methods are based on deep learning,but there are still many shortcomings in these methods,such as the problem of overfitting is easy to occur,and the problem of unbalanced data distribution has not been well solved.In this paper,biomedical named entity recognition based on reinforcement learning is proposed.Based on the traditional deep learning method,the context semantic information of words is learned by Bi-directional Long Short-Term Memory neural network.And the hidden layer state of its output is taken as the input of reinforcement learning,the Deep Q Learning algorithm is used to generate the annotation of the sequence and complete the decoding of tags.Compared with using the traditional feedforward neural network and conditional random field to decode,the method of reinforcement learning model can learn the long distance features better.In addition,by setting an appropriate reward mechanism,the interference caused by data noise or unbalanced samples in supervised learning can be effectively avoided.Finally,the experiment proves that the method based on reinforcement learning can achieve excellent results in biomedical entity recognition.This paper also combines theory with practice,we develops a disease knowledge extraction system by applying biomedical entity recognition algorithm model.User interface can be viewed through the system related to neurodegenerative disease symptom,chemical,gene,peptide,natural product,such as biomedical entities,the researchers in medical field to provide a knowledge graph visualization system,aided by the method of constructing network medical experts to study related diseases.