HIT-SCIR at MMNLU-22: Consistency Regularization for Multilingual Spoken Language Understanding

TL;DR

This paper introduces a consistency regularization approach with hybrid data augmentation to enhance multilingual spoken language understanding, significantly improving intent detection and slot filling performance, and achieving top results in a major competition.

Contribution

It proposes a novel consistency regularization method with hybrid data augmentation for multilingual SLU, leading to state-of-the-art results.

Findings

Improved intent detection accuracy.

Enhanced slot filling performance.

Ranked 1st in MMNLU-22 competition.

Abstract

Multilingual spoken language understanding (SLU) consists of two sub-tasks, namely intent detection and slot filling. To improve the performance of these two sub-tasks, we propose to use consistency regularization based on a hybrid data augmentation strategy. The consistency regularization enforces the predicted distributions for an example and its semantically equivalent augmentation to be consistent. We conduct experiments on the MASSIVE dataset under both full-dataset and zero-shot settings. Experimental results demonstrate that our proposed method improves the performance on both intent detection and slot filling tasks. Our system\footnote{The code will be available at \url{https://github.com/bozheng-hit/MMNLU-22-HIT-SCIR}.} ranked 1st in the MMNLU-22 competition under the full-dataset setting.