Simultaneously Linking Entities and Extracting Relations from Biomedical Text without Mention-Level Supervision
Authors: Trapit Bansal, Pat Verga, Neha Choudhary, Andrew McCallum7407-7414
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show that our model outperforms a state-of-the art entity linking and relation extraction pipeline on two biomedical datasets and can drastically improve the overall recall of the system. |
| Researcher Affiliation | Collaboration | University of Massachusetts, Amherst {tbansal, nchoudhary, mccallum}@cs.umass.edu Google Research patverga@google.com |
| Pseudocode | No | The paper describes the model architecture and mathematical formulations, but it does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code and data: https://github.com/theTB/snerl |
| Open Datasets | Yes | Our first set of experiments are on the CTD dataset first introduced in Verga, Strubell, and Mc Callum (2018). The data is derived from annotations in the Chemical Toxicology Database (Davis et al. 2018)... |
| Dataset Splits | Yes | We remedy this and create a more challenging train/development/test split from the entire CTD annotations... We consider k as a hyperparameter and tune it on the validation set. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for conducting the experiments. |
| Software Dependencies | No | The paper mentions using a Transformer architecture, Bio Sent Vec, and Dist Mult, but it does not specify version numbers for these or other software dependencies necessary for replication. |
| Experiment Setup | Yes | We train based on the cross-entropy loss... We consider k as a hyperparameter and tune it on the validation set. In summary, we combine graph prediction and document-level entity prediction objectives similar to multitask learning (Caruana 1993). |