The paper explores how to enhance biomedical language models using adapter modules and knowledge graphs. The approach involves partitioning knowledge graphs into smaller subgraphs, fine-tuning adapter modules for each subgraph, and combining the knowledge in a fusion layer. The methodology was tested on document classification, question answering, and natural language inference tasks, showing performance improvements in several instances. The approach is beneficial as it requires low computing power, making it feasible for smaller research groups.
Publication date: 22 Dec 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2312.13881
