All terms
Training
Domain Adaptation
Adjusting a model so it performs better on a specific subject area or data distribution.
Definition
Domain adaptation adjusts a model so it works better in a specific area, such as legal, medical, or financial text, that differs from the data it was originally trained on. It is commonly done through continued pretraining or fine-tuning on a targeted corpus, which teaches domain vocabulary and conventions. The aim is stronger performance on the target domain without retraining from scratch.