Re-train or Train from Scratch? Comparing Pre-training Strategies of BERT in the Medical Domain

Hicham El Boukkouri (1, 2) , Olivier Ferret (3) , Thomas Lavergne (1, 2) , Pierre Zweigenbaum (1, 2)

1 LISN - Laboratoire Interdisciplinaire des Sciences du Numérique
2 ILES - Information, Langue Ecrite et Signée
3 DIASI - Département Intelligence Ambiante et Systèmes Interactifs

BERT models used in specialized domains all seem to be the result of a simple strategy: initializing with the original BERT then resuming pre-training on a specialized corpus. This method yields rather good performance (e.g. BioBERT (Lee et al., 2020), SciBERT (Beltagy et al., 2019), BlueBERT (Peng et al., 2019)). However, it seems reasonable to think that training directly on a specialized corpus, using a specialized vocabulary, could result in more tailored embeddings and thus help performance. To test this hypothesis, we train BERT models from scratch using many configurations involving general and medical corpora. Based on evaluations using four different tasks, we find that the initial corpus only has a weak influence on the performance of BERT models when these are further pre-trained on a medical corpus.

Domains: Document and Text Processing

Hicham El Boukkouri, Olivier Ferret, Thomas Lavergne, Pierre Zweigenbaum. Re-train or Train from Scratch? Comparing Pre-training Strategies of BERT in the Medical Domain. LREC 2022, Jun 2022, Marseille, France. pp.2626-2633, 2022, LREC 2022. ⟨cea-03745341⟩