Lightly supervised vs. semi-supervised training of acoustic model on Luxembourgish for low-resource automatic speech recognition

Lightly supervised vs. semi-supervised training of acoustic model on Luxembourgish for low-resource automatic speech recognition

Paper in proceedings (conference paper)

In this work, we focus on exploiting inexpensive data in orderto to improve the DNN acoustic model for ASR. We exploretwo strategies: The first one uses untranscribed data fromthe target domain. The second one is related to the proper selectionof excerpts from imperfectly transcribed out-of-domainpublic data, as parliamentary speeches. We found out that bothapproaches lead to similar results, making them equally beneficialfor practical use. The Luxembourgish ASR seed systemhad a 38.8% WER and it improved by roughly 4% absolute,leading to 34.6% for untranscribed and 34.9% for lightlysuperviseddata. Adding both databases simultaneously ledto 34.4% WER, which is only a small improvement. As asecondary research topic, we experiment with semi-supervisedstate-level minimum Bayes risk (sMBR) training. Nonetheless,for sMBR we saw no improvement from adding the automaticallytranscribed target data, despite that similar techniquesyield good results in the case of cross-entropy (CE) training.

In this work, we focus on exploiting inexpensive data in orderto to improve the DNN acoustic model for ASR. We exploretwo strategies: The first one uses untranscribed data fromthe target domain. The second one is related to the proper selectionof excerpts from imperfectly transcribed out-of-domainpublic data, as parliamentary speeches. We found out that bothapproaches lead to similar results, making them equally beneficialfor practical use. The Luxembourgish ASR seed systemhad a 38.8% WER and it improved by roughly 4% absolute,leading to 34.6% for untranscribed and 34.9% for lightlysuperviseddata. Adding both databases simultaneously ledto 34.4% WER, which is only a small improvement. As asecondary research topic, we experiment with semi-supervisedstate-level minimum Bayes risk (sMBR) training. Nonetheless,for sMBR we saw no improvement from adding the automaticallytranscribed target data, despite that similar techniquesyield good results in the case of cross-entropy (CE) training.

Luxembourgish, call centers, speech recognition,low-resourced ASR, unsupervised training

Luxembourgish, call centers, speech recognition,low-resourced ASR, unsupervised training

VESELÝ, K.; PERALES, C.; SZŐKE, I.; LUQUE, J.; ČERNOCKÝ, J.

2019

02.09.2018

International Speech Communication Association

Hyderabad

Proceedings of Interspeech 2018

1990-9772

Proceedings of Interspeech

2018

9

French Republic

2883

2887

5

https://www.isca-speech.org/archive/Interspeech_2018/abstracts/2361.html

vesely_interspeech2018_2361