Cross lingual transfer learning for zero-resource domain adaptation

TL;DR

This paper introduces a cross-lingual transfer learning method for zero-resource domain adaptation of DNN acoustic models, enabling effective domain adaptation in low-resource languages by leveraging multi-lingual shared layers and well-resourced language data.

Contribution

The paper presents a novel multi-lingual DNN architecture that allows domain adaptation transforms from a well-resourced language to be applied to low-resource languages, improving speech recognition performance.

Findings

29% relative WER improvement on Spanish BN test data using English adaptation data

18-27% relative WER improvements for low-resource languages with poor language match

Outperforms multi-task and multi-condition training approaches in domain adaptation

Abstract

We propose a method for zero-resource domain adaptation of DNN acoustic models, for use in low-resource situations where the only in-language training data available may be poorly matched to the intended target domain. Our method uses a multi-lingual model in which several DNN layers are shared between languages. This architecture enables domain adaptation transforms learned for one well-resourced language to be applied to an entirely different low-resource language. First, to develop the technique we use English as a well-resourced language and take Spanish to mimic a low-resource language. Experiments in domain adaptation between the conversational telephone speech (CTS) domain and broadcast news (BN) domain demonstrate a 29% relative WER improvement on Spanish BN test data by using only English adaptation data. Second, we demonstrate the effectiveness of the method for low-resource languages with a poor match to the well-resourced language. Even in this scenario, the proposed method achieves relative WER improvements of 18-27% by using solely English data for domain adaptation. Compared to other related approaches based on multi-task and multi-condition training, the proposed method is able to better exploit well-resource language data for improved acoustic modelling of the low-resource target domain.