Sequence-to-Sequence Learning with Latent Neural Grammars

TL;DR

This paper introduces a hierarchical sequence-to-sequence model using latent neural grammars, which induces source and target trees during training, aiming to improve compositional generalization and performance on various tasks.

Contribution

It proposes a novel neural grammar-based approach for sequence-to-sequence learning with latent trees, enhancing compositional generalization without manual feature engineering.

Findings

Performs well on compositional generalization tasks like SCAN

Effective in style transfer and small-scale translation

Outperforms standard baselines in tested domains

Abstract

Sequence-to-sequence learning with neural networks has become the de facto standard for sequence prediction tasks. This approach typically models the local distribution over the next word with a powerful neural network that can condition on arbitrary context. While flexible and performant, these models often require large datasets for training and can fail spectacularly on benchmarks designed to test for compositional generalization. This work explores an alternative, hierarchical approach to sequence-to-sequence learning with quasi-synchronous grammars, where each node in the target tree is transduced by a node in the source tree. Both the source and target trees are treated as latent and induced during training. We develop a neural parameterization of the grammar which enables parameter sharing over the combinatorial space of derivation rules without the need for manual feature engineering. We apply this latent neural grammar to various domains -- a diagnostic language navigation task designed to test for compositional generalization (SCAN), style transfer, and small-scale machine translation -- and find that it performs respectably compared to standard baselines.