Improving Code-switching Language Modeling with Artificially Generated Texts using Cycle-consistent Adversarial Networks

TL;DR

This paper introduces a cycle-consistent adversarial network framework to generate artificial code-switching texts from monolingual data, enhancing language models and speech recognition in low-resource scenarios.

Contribution

It proposes a novel GAN-based method to artificially generate code-switching data, addressing data scarcity in code-switching language modeling.

Findings

Artificially generated data improves language model performance.

Enhanced speech recognition accuracy with generated code-switching texts.

Method effective on SEAME corpus.

Abstract

This paper presents our latest effort on improving Code-switching language models that suffer from data scarcity. We investigate methods to augment Code-switching training text data by artificially generating them. Concretely, we propose a cycle-consistent adversarial networks based framework to transfer monolingual text into Code-switching text, considering Code-switching as a speaking style. Our experimental results on the SEAME corpus show that utilising artificially generated Code-switching text data improves consistently the language model as well as the automatic speech recognition performance.