Joint learning-based causal relation extraction from biomedical literature.

Area of Science:

• Biomedical text mining
• Natural Language Processing
• Bioinformatics

Background:

• Causal relation extraction in biomedical text is challenging, requiring identification of both entity relations and functions.
• Independent learning for relation extraction and function detection overlooks their intrinsic correlation, limiting performance.
• Existing methods often treat relation extraction and function detection as separate tasks, hindering overall accuracy.

Purpose of the Study:

• To develop a joint learning model that integrates entity relation extraction and entity function detection.
• To improve the performance of biomedical causal relation extraction by exploiting the inter-relationship between entities and their functions.
• To achieve state-of-the-art results in BEL (Biomedical Event and Relation) statement extraction.

Main Methods:

• Proposed a novel joint learning model combining entity relation extraction and entity function detection.
• Trained and evaluated the model on the BioCreative-V Track 4 corpus.
• Compared the joint learning model against separate learning approaches.

Main Results:

• The joint learning model significantly outperformed separate models in BEL statement extraction.
• Achieved F1 scores of 57.0% (Stage 2) and 37.3% (Stage 1) on the test set.
• Demonstrated state-of-the-art performance in Stage 2 compared to other systems.

Conclusions:

• Joint learning effectively captures the inter-relationship between entity relations and functions.
• The proposed model enhances the accuracy of biomedical causal relation extraction.
• The system achieves competitive and state-of-the-art performance on benchmark datasets.