Enhancing Entity Representations with Prompt Learning for Biomedical Entity Linking

Authors: Tiantian Zhu, Yang Qin, Qingcai Chen, Baotian Hu, Yang Xiang

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that our model achieves promising performance improvements compared with several state-of-the-art techniques on the largest biomedical public dataset Med Mentions and the NCBI disease corpus. Finally, we evaluate the performance of our model using the Med Mentions [Mohan and Li, 2019] and the NCBI disease [Dogan et al., 2014] datasets, and against several state-of-the-art (SOTA) baselines. 4 Experiments
Researcher Affiliation Academia 1Harbin Institute of Technology (Shenzhen), Shenzhen, China 2Peng Cheng Laboratory, Shenzhen, China zhu.tiantian110@gmail.com, {csyqin, qingcai.chen, hubaotian}@hit.edu.cn, xiangy@pcl.ac.cn
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks. The method is described textually and with mathematical equations.
Open Source Code Yes The code is available at https://github.com/Tiantian Zhu110/Bio PRO.
Open Datasets Yes We used two datasets to evaluate our proposed model: Med Mentions [Mohan and Li, 2019] and NCBI Disease [Dogan et al., 2014] datasets.
Dataset Splits Yes Table 1: Statistics of different sets in Med Mentions and NCBI Disease datasets. Med Mentions Documents 2, 635 878 879 Mentions 211, 029 71, 062 70, 405 Entities 25, 640 12, 586 12, 402 NCBI Disease Documents 592 100 100 Mentions 5, 134 787 960 Entities 668 176 203
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions using pre-trained SAPBERT and Pub Med BERT, and that the optimizer is Adam W, but it does not provide specific version numbers for these software components (e.g., PyTorch version, specific BERT/SAPBERT implementation versions).
Experiment Setup Yes Table 2: The search space for hyper-parameters used in prompt-tuning. denotes the used ones for reporting results. Optimizer Adam W Learning Rate 1e 4, 1e 5, 2e 6, 1e 6 Batch Size 128, 64, 32 Epoch 15 Max Sequence Length 256 Soft Token Embedding Size 768 Weight Decay 0.01