Multilingual End-to-End Speech Recognition with A Single Transformer on Low-Resource Languages

TL;DR

This paper demonstrates that a single Transformer model can effectively perform multilingual low-resource speech recognition using sub-words without pronunciation lexicons, and that incorporating language information improves accuracy.

Contribution

It introduces a multilingual Transformer-based ASR model for low-resource languages that employs sub-words and integrates language info to enhance recognition performance.

Findings

Single Transformer performs well on low-resource languages despite language confusion.

Inserting language symbols at sequence ends yields better WER reduction.

Language information inclusion leads to approximately 10.5-12.4% WER improvement.

Abstract

Sequence-to-sequence attention-based models integrate an acoustic, pronunciation and language model into a single neural network, which make them very suitable for multilingual automatic speech recognition (ASR). In this paper, we are concerned with multilingual speech recognition on low-resource languages by a single Transformer, one of sequence-to-sequence attention-based models. Sub-words are employed as the multilingual modeling unit without using any pronunciation lexicon. First, we show that a single multilingual ASR Transformer performs well on low-resource languages despite of some language confusion. We then look at incorporating language information into the model by inserting the language symbol at the beginning or at the end of the original sub-words sequence under the condition of language information being known during training. Experiments on CALLHOME datasets demonstrate that the multilingual ASR Transformer with the language symbol at the end performs better and can obtain relatively 10.5\% average word error rate (WER) reduction compared to SHL-MLSTM with residual learning. We go on to show that, assuming the language information being known during training and testing, about relatively 12.4\% average WER reduction can be observed compared to SHL-MLSTM with residual learning through giving the language symbol as the sentence start token.