Related Experiment Video
Updated: Jul 17, 2026
07:35
A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports
Published on: October 13, 2023
Domain adaptation for semantic role labeling in the biomedical domain
Daniel Dahlmeier1, Hwee Tou Ng
1NUS Graduate School for Integrative Sciences and Engineering, Singapore 117456, Singapore.
Bioinformatics (Oxford, England)
|February 25, 2010
Summary
Domain adaptation significantly reduces the cost of developing biomedical semantic role labeling (SRL) systems. This approach achieves high performance with limited annotated biomedical data, making SRL more accessible.
Area of Science:
- Natural Language Processing (NLP)
- Biomedical Informatics
Background:
- Semantic Role Labeling (SRL) is crucial for extracting meaning from text.
- Existing biomedical SRL systems require extensive manual annotation, which is costly and time-consuming.
- Leveraging existing resources from other domains is a key challenge.
Purpose of the Study:
- To address the scarcity of annotated biomedical data for SRL.
- To adapt existing SRL resources from the newswire domain to the biomedical domain.
- To reduce the cost and effort in developing biomedical SRL systems.
Main Methods:
- Framing biomedical SRL as a domain adaptation problem.
- Evaluating three domain adaptation algorithms for SRL.
- Utilizing the BioKIT system for biomedical SRL.
Main Results:
- Domain adaptation significantly reduces the cost of developing biomedical SRL systems.
- The proposed approach achieves 97% of the performance with only 60 annotated abstracts.
- Demonstrates the effectiveness of domain adaptation in overcoming data scarcity.
Conclusions:
- Domain adaptation is a viable and cost-effective strategy for biomedical SRL.
- It enables the development of high-performance SRL systems with substantially less annotated data.
- Facilitates broader application of SRL in biomedical text analysis.
