Learning to Generalize Compositionally by Transferring Across Semantic Parsing Tasks
This addresses the issue of domain mismatch in NLP for researchers and practitioners, though it is incremental as it builds on existing transfer learning techniques.
The paper tackles the problem of poor compositional generalization in neural networks by developing a method that learns transferable representations across semantic parsing tasks, resulting in significant improvements on target task test sets.
Neural network models often generalize poorly to mismatched domains or distributions. In NLP, this issue arises in particular when models are expected to generalize compositionally, that is, to novel combinations of familiar words and constructions. We investigate learning representations that facilitate transfer learning from one compositional task to another: the representation and the task-specific layers of the models are strategically trained differently on a pre-finetuning task such that they generalize well on mismatched splits that require compositionality. We apply this method to semantic parsing, using three very different datasets, COGS, GeoQuery and SCAN, used alternately as the pre-finetuning and target task. Our method significantly improves compositional generalization over baselines on the test set of the target task, which is held out during fine-tuning. Ablation studies characterize the utility of the major steps in the proposed algorithm and support our hypothesis.