Out-of-distribution generalisation in spoken language understanding

TL;DR

This paper introduces a new dataset split for evaluating out-of-distribution generalisation in spoken language understanding, revealing current models' limited ability to generalise and highlighting the need for novel techniques.

Contribution

It presents SLURPFOOD, a modified SLU dataset with OOD splits, and analyses the generalisation challenges faced by models using interpretability methods.

Findings

End-to-end SLU models show limited OOD generalisation.

Model interpretability reveals factors affecting generalisation.

Two techniques improve some OOD split performances.

Abstract

Test data is said to be out-of-distribution (OOD) when it unexpectedly differs from the training data, a common challenge in real-world use cases of machine learning. Although OOD generalisation has gained interest in recent years, few works have focused on OOD generalisation in spoken language understanding (SLU) tasks. To facilitate research on this topic, we introduce a modified version of the popular SLU dataset SLURP, featuring data splits for testing OOD generalisation in the SLU task. We call our modified dataset SLURP For OOD generalisation, or SLURPFOOD. Utilising our OOD data splits, we find end-to-end SLU models to have limited capacity for generalisation. Furthermore, by employing model interpretability techniques, we shed light on the factors contributing to the generalisation difficulties of the models. To improve the generalisation, we experiment with two techniques, which improve the results on some, but not all the splits, emphasising the need for new techniques.