Source-Free Domain Adaptation for Question Answering with Masked Self-training

arXiv:2212.09563v2 Announce Type: replace
Abstract: Most previous unsupervised domain adaptation (UDA) methods for question answering(QA) require access to source domain data while fine-tuning the model for the target domain. Source domain data may, however, contain sensitive information and may be restricted. In this study, we investigate a more challenging setting, source-free UDA, in which we have only the pretrained source model and target domain data, without access to source domain data. We propose a novel self-training approach to QA models …

abstract arxiv cs.cl data domain domain adaptation fine-tuning free however information question question answering self-training study training type unsupervised