Revisiting Iterative Back-Translation from the Perspective of Compositional Generalization
arXiv:2012.04276
Abstract
Human intelligence exhibits compositional generalization (i.e., the capacity to understand and produce unseen combinations of seen components), but current neural seq2seq models lack such ability. In this paper, we revisit iterative back-translation, a simple yet effective semi-supervised method, to investigate whether and how it can improve compositional generalization. In this work: (1) We first empirically show that iterative back-translation substantially improves the performance on compositional generalization benchmarks (CFQ and SCAN). (2) To understand why iterative back-translation is useful, we carefully examine the performance gains and find that iterative back-translation can increasingly correct errors in pseudo-parallel data. (3) To further encourage this mechanism, we propose curriculum iterative back-translation, which better improves the quality of pseudo-parallel data, thus further improving the performance.
accepted in AAAI 2021
References in corpus (7)
- Dual Learning for Machine Translation
- Compositional generalization in a deep seq2seq model by separating syntax and semantics
- Compositional Generalization in Semantic Parsing: Pre-training vs. Specialized Architectures
- Measuring Compositional Generalization: A Comprehensive Method on Realistic Data
- Learning Compositional Rules via Neural Program Synthesis
- Siamese recurrent networks learn first-order logic reasoning and exhibit zero-shot compositional generalization
- Explaining and Generalizing Back-Translation through Wake-Sleep