Unlock instant, AI-driven research and patent intelligence for your innovation.

Prescription Extraction Method of Ancient Medical Records Based on Hierarchical Sequence Labeling

A sequence labeling and prescription technology, applied in the field of pre-training language model, can solve the problems of rough classification, error propagation, error, etc., achieve the effect of enhancing feature representation, good auxiliary function, and reducing labeling pressure

Active Publication Date: 2021-08-27
ZHEJIANG UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

There are two problems with this method: (1) Sentence clauses are based on rules, which is prone to errors and lead to error propagation; (2) Sentence-based classification is too rough, and some prescriptions are only part of the sentence, rather than the whole The sentences are all prescriptions

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Prescription Extraction Method of Ancient Medical Records Based on Hierarchical Sequence Labeling
  • Prescription Extraction Method of Ancient Medical Records Based on Hierarchical Sequence Labeling
  • Prescription Extraction Method of Ancient Medical Records Based on Hierarchical Sequence Labeling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] The present invention will be described in detail below in conjunction with specific examples.

[0030] In the current information extraction task, the goal is mainly focused on how to extract the named entity of the text, and the named entity is usually very short. Compared with the extraction of prescription text, the prescription text is usually a long sequence of medication. For this reason, the present invention proposes two methods, one is to extract prescriptions based on hierarchical sequence labeling. Treat prescription extraction as a sequence labeling problem, and use the BIO labeling system to mark the corresponding fragments of the prescription. The other is a method based on boundary prediction, which predicts the start and end positions of prescription fragments in the text. Through a large number of experiments, we have proved that the scheme based on hierarchical sequence annotation is better than the method based on boundary prediction, so we finally ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a method for extracting prescriptions from ancient medical cases based on hierarchical sequence annotation, which adopts the layered sequence annotation network of BERT+CRF, including input layer, feature extraction layer, fully connected layer, medicine and prescription name prediction CRF layer, and prescription prediction CRF layer. First of all, the present invention does not need to divide the medical case into sentences, and directly takes the complete medical case as input, avoiding the error propagation caused by the sentence division. Second, the prescription text is obtained in the form of sequence annotations to directly obtain the most relevant text fragments. Finally, the information of the drug name and prescription name is considered in the recognition process, which enhances the feature representation in the prescription extraction process and achieves better results. A small amount of manually labeled data can be used to identify the prescription text in ancient medical records; The invention also designs a BLEU-based evaluation index method suitable for model selection to quantify the matching level between model extraction results and labeling results to obtain the best model.

Description

technical field [0001] The invention relates to a pre-trained language model in deep learning and a conditional random field. Specifically, it is a method for extracting prescriptions from ancient medical records based on hierarchical sequence annotation. Background technique [0002] TCM medical records record the complete process of a patient's treatment of a disease, including symptoms and prescriptions during treatment, etc. However, due to reasons such as the era in which ancient doctors lived and their personal styles, the content of ancient medical records was complicated and their formats varied greatly. This has brought difficulties to the formatting and digitization of medical record content. For Chinese medicine scholars and lovers of Chinese medicine, learning the treatment experience of past doctors from medical records is an important way to learn treatment ideas. For this reason, how to format the text of ancient medical records is particularly important. ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F40/211G06F40/279G16H20/10G16H50/70
CPCG16H20/10G16H50/70
Inventor 张引熊海辉
Owner ZHEJIANG UNIV