A Two-Stage Framework with Self-Supervised Distillation For Cross-Domain Text Classification
This addresses the problem of adapting text classification models to new domains with limited labeled data, representing an incremental advance in domain adaptation methods.
The paper tackles cross-domain text classification by proposing a two-stage framework with self-supervised distillation, achieving state-of-the-art results with improvements of 1.03% to 94.17% for single-source and 1.34% to 95.09% for multi-source adaptations.
Cross-domain text classification aims to adapt models to a target domain that lacks labeled data. It leverages or reuses rich labeled data from the different but related source domain(s) and unlabeled data from the target domain. To this end, previous work focuses on either extracting domain-invariant features or task-agnostic features, ignoring domain-aware features that may be present in the target domain and could be useful for the downstream task. In this paper, we propose a two-stage framework for cross-domain text classification. In the first stage, we finetune the model with mask language modeling (MLM) and labeled data from the source domain. In the second stage, we further fine-tune the model with self-supervised distillation (SSD) and unlabeled data from the target domain. We evaluate its performance on a public cross-domain text classification benchmark and the experiment results show that our method achieves new state-of-the-art results for both single-source domain adaptations (94.17% $\uparrow$1.03%) and multi-source domain adaptations (95.09% $\uparrow$1.34%).