Transformers and the Representation of Biomedical Background Knowledge

Oskar Wysocki, Zili Zhou, Paul O’Regan, Deborah Ferreira, Magdalena Wysocka, Dónal Landers, André Freitas


Abstract
Specialized transformers-based models (such as BioBERT and BioMegatron) are adapted for the biomedical domain based on publicly available biomedical corpora. As such, they have the potential to encode large-scale biological knowledge. We investigate the encoding and representation of biological knowledge in these models, and its potential utility to support inference in cancer precision medicine—namely, the interpretation of the clinical significance of genomic alterations. We compare the performance of different transformer baselines; we use probing to determine the consistency of encodings for distinct entities; and we use clustering methods to compare and contrast the internal properties of the embeddings for genes, variants, drugs, and diseases. We show that these models do indeed encode biological knowledge, although some of this is lost in fine-tuning for specific tasks. Finally, we analyze how the models behave with regard to biases and imbalances in the dataset.
Anthology ID:
2023.cl-1.2
Volume:
Computational Linguistics, Volume 49, Issue 1 - March 2023
Month:
March
Year:
2023
Address:
Cambridge, MA
Venue:
CL
SIG:
Publisher:
MIT Press
Note:
Pages:
73–115
Language:
URL:
https://aclanthology.org/2023.cl-1.2
DOI:
10.1162/coli_a_00462
Bibkey:
Cite (ACL):
Oskar Wysocki, Zili Zhou, Paul O’Regan, Deborah Ferreira, Magdalena Wysocka, Dónal Landers, and André Freitas. 2023. Transformers and the Representation of Biomedical Background Knowledge. Computational Linguistics, 49(1):73–115.
Cite (Informal):
Transformers and the Representation of Biomedical Background Knowledge (Wysocki et al., CL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.cl-1.2.pdf