Data augmentation using prosody and false starts to recognize non-native children's speech

TL;DR

This paper presents a data augmentation approach using prosody and false start simulation to improve speech recognition accuracy for non-native children's spontaneous speech, achieving a notable reduction in word error rate.

Contribution

It introduces a novel combination of prosody-based data augmentation and false start simulation to enhance ASR performance on non-native children's speech.

Findings

Prosody-based augmentation outperforms baseline and SpecAugment.

Simulating false starts improves language model accuracy.

Combined methods reduce WER to 17.99%.

Abstract

This paper describes AaltoASR's speech recognition system for the INTERSPEECH 2020 shared task on Automatic Speech Recognition (ASR) for non-native children's speech. The task is to recognize non-native speech from children of various age groups given a limited amount of speech. Moreover, the speech being spontaneous has false starts transcribed as partial words, which in the test transcriptions leads to unseen partial words. To cope with these two challenges, we investigate a data augmentation-based approach. Firstly, we apply the prosody-based data augmentation to supplement the audio data. Secondly, we simulate false starts by introducing partial-word noise in the language modeling corpora creating new words. Acoustic models trained on prosody-based augmented data outperform the models using the baseline recipe or the SpecAugment-based augmentation. The partial-word noise also helps to improve the baseline language model. Our ASR system, a combination of these schemes, is placed third in the evaluation period and achieves the word error rate of 18.71%. Post-evaluation period, we observe that increasing the amounts of prosody-based augmented data leads to better performance. Furthermore, removing low-confidence-score words from hypotheses can lead to further gains. These two improvements lower the ASR error rate to 17.99%.