A tag based joint extraction model for Chinese medical text

Area of Science:

• Natural Language Processing
• Medical Informatics
• Computational Linguistics

Background:

• Information extraction structures medical knowledge and aids discovery.
• Traditional pipelined methods neglect entity and relation extraction relevance.
• Joint extraction models show promise but are under-explored in medical fields.

Purpose of the Study:

• To develop a joint extraction model for Chinese medical texts.
• To address limitations of traditional pipelined approaches.
• To improve the accuracy of medical information extraction.

Main Methods:

• Pretreatment of Chinese medical data for word sequences.
• A BIOH12D1D2 tagging scheme for joint extraction and overlapping entities.
• An encoder-decoder model with BERT for enhanced feature representation.

Main Results:

• The joint extraction model achieved an F1 score of 0.7 on the CHIP-2020 dataset.
• This represents a 0.364 improvement compared to the baseline.
• The model effectively handles Chinese medical text information extraction.

Conclusions:

• The proposed joint extraction model enhances medical knowledge structuring.
• The BIOH12D1D2 tagging scheme and BERT integration are effective.
• This work contributes to advancing information extraction in the medical domain.