Deriving Machine Attention from Human Rationales
This addresses the challenge of data scarcity for attention-based models, offering a transferable solution across domains.
The paper tackles the problem of learning attention in low-resource scenarios by mapping discrete human-annotated rationales into continuous attention, achieving over 15% average error reduction on benchmark datasets.
Attention-based models are successful when trained on large amounts of data. In this paper, we demonstrate that even in the low-resource scenario, attention can be learned effectively. To this end, we start with discrete human-annotated rationales and map them into continuous attention. Our central hypothesis is that this mapping is general across domains, and thus can be transferred from resource-rich domains to low-resource ones. Our model jointly learns a domain-invariant representation and induces the desired mapping between rationales and attention. Our empirical results validate this hypothesis and show that our approach delivers significant gains over state-of-the-art baselines, yielding over 15% average error reduction on benchmark datasets.