go to ORKG: http://orkg.org/orkg/predicate/P44186

Pre-train then Fine-tune

We first pre-train BERT on the source domain data, and then fine-tune it to the target domain samples.