A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT

Renlong QI

doi:10.12184/taposatWSP2633-456904.20200101

PDF
Export
Share
Collection
Album

A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT

Volume 1, Issue 1

Views: 0 Downloads: 0

A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT
A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT
1. Zhengzhou University of Science and Technology
2. College of Information Science and Engineering, Henan University of Technology
Vol. 1, Issue 1, Pages: 19-31(2020)
DOI：10.12184/taposatWSP2633-456904.20200101
Accepted：

Scan QR Code

RENLONG QI. (2020). A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT. Theory and Practice of Science and Technology, 1(1), 19-31..
DOI：

RENLONG QI. (2020). A Named Entity Recognition Method for Chinese Electronic Medical Records Based on BERT. Theory and Practice of Science and Technology, 1(1), 19-31.. DOI： 10.12184/taposatWSP2633-456904.20200101.

摘要

Abstract

When Chinese electronic medical records is processed

it is still a challenge to distinguish the multiple manifestations for one entity with traditional Named Entity Recognition model. In this paper

a novel NER scheme for Chinese EMRs based on BERT is proposed. Firstly

the semantic vector of characters is generated dynamically by BERT; Secondly

the generated word vectors is inputed into BiLSTM model to obtain the global features of the text; Finally

the Conditional Random Field model is used to decode the global features of the text to generate the optimal annotation sequence. Experiments show that the proposed model achieves a recall rate of 96.35% on the NER of EMRs

which is better than the traditional recognition method.

关键词

Keywords

Electronic Medical RecordBERTNERCRF-BiLSTM model

references

Alert me when the article has been cited

Submit

Sentiment Analysis of Airline Reviews:Based on BERT-CNN Model

Related Author

Pei Gan

Related Institution

School of Computer Science & Engineering, UESTC

⁰