GraphPrompt: Graph-Based Prompt Templates for Biomedical Synonym Prediction
Authors: Hanwen Xu, Jiayou Zhang, Zhirui Wang, Shizhuo Zhang, Megh Bhalerao, Yucong Liu, Dawei Zhu, Sheng Wang
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on the novel OBO-syn dataset demonstrate the superior performance of our method against existing synonym prediction approaches, indicating the advantage of considering the graph structure. We selected five graphs (mp, cl, hp, fbbt, doid) with the number of entities between 10,000 and 20,000 from OBO-syn. We investigated a few-shot setting and a zero-shot setting. |
| Researcher Affiliation | Academia | 1University of Washington, 2Mohamed bin Zayed University of Artificial Intelligence 3Carnegie Mellon University, 4Nanyang Technological University, 5Peking University |
| Pseudocode | No | The paper describes its methods through prose and mathematical equations but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | All the data and codes are avalible at: https://github.com/Hanwen Xu THU/Graph Prompt |
| Open Datasets | Yes | We collected 70 relational graphs from Open Biological and Biomedical Ontology Foundry (OBO) (Smith et al. 2007). All the data and codes are avalible at: https://github.com/Hanwen Xu THU/Graph Prompt |
| Dataset Splits | Yes | In the few-shot setting, we split the synonyms into six folds, and then used four folds as training set, one fold as validation set and one fold as test set. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for experiments, such as GPU models, CPU types, or memory. |
| Software Dependencies | No | The paper mentions using 'Bio BERT' and 'Sentence BERT' but does not specify version numbers for these or any other software libraries, frameworks (e.g., PyTorch, TensorFlow), or programming languages used. |
| Experiment Setup | Yes | We selected Bio BERT with [CLS] readout function as Encv and Encs, and share the parameters between both encoders. To alleviate this issue, we use a trainable 1-d Batch Norm layer and redefine our similarity function Q as: Q(xv, xs) = exp(BN(x T v xs)). Besides, we utilize the hard negative strategy following Sung et al. (2020) by sampling difficult negative candidates U V. The possible choices of these weights are 2, 3 for Lp, 0 and 1 for Lc1, 0 and 1 for Lc2, 1 and 2 for λ1 and λ2. We used grid search to find the best combinations of these values. |