Constrained Output Embeddings for End-to-End Code-Switching Speech Recognition with Only Monolingual Data

Abstract

The lack of code-switch training data is one of the major concerns in the development of end-to-end code-switching automatic speech recognition (ASR) models. In this work, we propose a method to train an improved end-to-end code-switching ASR using only monolingual data. Our method encourages the distributions of output token embeddings of monolingual languages to be similar, and hence, promotes the ASR model to easily code-switch between languages. Specifically, we propose to use Jensen-Shannon divergence and cosine distance based constraints. The former will enforce output embeddings of monolingual languages to possess similar distributions, while the later simply brings the centroids of two distributions to be close to each other. Experimental results demonstrate high effectiveness of the proposed method, yielding up to 4.5% absolute mixed error rate improvement on Mandarin-English code-switching ASR task.

5 pages, 3 figures, accepted to INTERSPEECH 2019

References in corpus (5)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Attention-Based Models for Speech Recognition
• Towards End-to-End Code-Switching Speech Recognition
• Learn to Code-Switch: Data Augmentation using Copy Mechanism on Language Modeling
• Code-switching Sentence Generation by Generative Adversarial Networks and its Application to Data Augmentation

Cited by in corpus (7)

• A Survey of Code-switched Speech and Language Processing
• Data Augmentation for End-to-end Code-switching Speech Recognition
• Decoupling Pronunciation and Language for End-to-end Code-switching Automatic Speech Recognition
• A Study of Multilingual End-to-End Speech Recognition for Kazakh, Russian, and English
• Approaches to Improving Recognition of Underrepresented Named Entities in Hybrid ASR Systems
• Training a code-switching language model with monolingual data
• Code Switching Language Model Using Monolingual Training Data