Improving Sentence Classification in Abstracts of Randomized Controlled Trial using Prompt Learning
Yan Hu, Yong Chen, Hua Xu · 2022 IEEE 10th International Conference on Healthcare Informatics (ICHI) · 2022
We proposed a Prompt Learning (PL) based Sentence Classification method to structure abstracts of Randomized Controlled Trial (RCT) articles. Five manually designed templates and four different BERT models were developed for PL and we evaluated them on a dataset of 20K PubMed RCT abstracts. We also compared them with the hierarchical sequential labeling network (HSLN) approach, which previously achieved the best performance in relevant work. Our results show that the PL-based method outperformed HSLN by 0.0138, achieving an F1 score of 0.9401 on the PubMed 20k dataset, indicating the promise of PL in this task.