Improving Code-switching Language Modeling with Artificially Generated Texts using Cycle-consistent Adversarial Networks
This addresses data scarcity for Code-switching language modeling, which is an incremental improvement in a domain-specific area.
The paper tackles the problem of data scarcity in Code-switching language modeling by artificially generating Code-switching text using a cycle-consistent adversarial networks framework, resulting in consistent improvements in language model and automatic speech recognition performance on the SEAME corpus.
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.