Chinese medical named entity recognition method and device integrating knowledge map and bert

Abstract

The invention discloses a Chinese medical named entity recognition method and device that integrates knowledge graph and BERT. The method includes: constructing a data model of the knowledge graph; crawling the corresponding content in the disease introduction page on the medical website, and storing the content in database, and export the data in the database as a json file; extract the medical entity and entity relationship from the json file according to the data model, and construct the medical knowledge map K; identify the medical entity contained in the input sentence, and According to the medical knowledge graph, the medical entities are filled into triplets ε=(w i ,r k ,w j ); Inject the triple into the sentence to generate a sentence tree; process the sentence tree through the soft position index in the embedding layer and the visual matrix in the visual layer to solve the knowledge noise that may be generated by introducing external knowledge The problem and the BERT model cannot handle the problem of graph-structured sentence trees. The invention can quickly and effectively extract the medical entities included in the text, and has high accuracy.

Description

technical field [0001] The present invention relates to the field of computer technology, and more specifically, to a Chinese medical named entity recognition method and device that integrates knowledge graphs and BERT. Background technique [0002] With the continuous acceleration of the construction of medical and health informatization, the medical industry has also ushered in its own era of big data. Massive medical data generally exist in text in an unstructured way, how to efficiently extract useful information from these texts has become the focus of modern medical research and development. [0003] As the basis of natural language processing, NER can automatically extract specific entities such as diseases and symptom names from medical texts, which can help people to be liberated from the tedious manual labeling and analysis work. However, there are still many problems and difficulties in NER for the Chinese medical field. Chinese does not have a natural separator...

AI Technical Summary

Problems solved by technology

Compared with the method based on dictionaries and rules, this method has a greater improvement in accuracy, but it does not take into account the context of the word well, ignoring the semantic information of the word, and there is still a certain degree of recognition accuracy. upside

Images