Named Entity Recognition In Chinese Medical Text Based On Lattice LSTM

Medical texts contain vital clinical information,which has attracted the attention of researchers.Named Entity Recognition is the basis of data mining and information processing,and an important part of fully exploiting and utilizing valuable information in medical texts.The Named Entity Recognition technology can accurately extract the information that people need from medical texts,then it can help medical workers to optimize clinical decision-making,support evidence-based medicine and disease surveillance,thereby improving the overall health care quality.This paper proposes a Lattice LSTM(Long Short-Term Memory)based algorithm for Named Entity Recognition in Chinese medical texts.The algorithm proposed in this paper optimizes the current Named Entity Recognition algorithm in Chinese medical text,which cannot simultaneously take the character sequence information and the problem of error transmission into account.Combining the character information and word information of the medical text sequence,the overall named entity recognition effect is improved by correctly identifying the text boundary of the named entity.In the word embedding layer,this paper uses a large number of medical texts and professional medical dictionary training word and character vector models.Using the word and character vector,the medical text information can be embedded into the algorithm model better.In order to verify the validity of the algorithm,the model was tested on the competition data and the first course record data of Sichuan Cancer Hospital.Two other classic algorithms,conditional random field(CRF)and LSTM-CRF,were compared with this algorithm,the result shows that the accuracy of this algorithm is 0.3% higher than the other two,whether on the competition data or real first course record data.It proves that our algorithm can obtain better recognition result in the field of Named Entity Recognition in Chinese medical text.Based on our algorithm,this paper designs and implements a system for managing and analyzing electronic medical record.In addition to storing and managing electronic medical records,the system can also convert the electronic medical record into a tree structure,which is easier to display.In order to facilitate the clinical research of medical workers,the system provides a function that can help medical workers to collect a large number of similar medical records,it is helpful in analyzing and discussing the symptoms of a certain disease.Besides,this system can help researchers label medical texts,which can be used to build a medical text tagging corpus.