An Efficient Method for Biomedical Entity Linking Based on Inter- and Intra-Entity Attention

Biomedical entity linking is an important research problem for many downstream tasks, such as biomedical intelligent question answering, information retrieval, and extraction. the task of mapping mentions in medical texts to standard entities a given knowledge base. Recently, BERT-based models have achieved state-of-the-art results on task. Although this type method effective, it brings challenges fine-tuning online services practical industries due large number model parameters long inference time. In addition, numerous surface variants mentions, difficult single matching module achieve good results. To address challenge, we propose efficient that integrates inter- intra-entity attention better capture between candidate themselves each other, paper more lightweight. Experimental show our achieves competitive performance two benchmark datasets, NCBI ADR, with accuracy rate 91.28% 93.13%, respectively. Moreover, also comparable or even compared while having far fewer very high speed.