Do end-to-end speech recognition models care about context?
Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Standard
Do end-to-end speech recognition models care about context? / Borgholt, Lasse; Havtorn, Jakob D.; Agic, Željko; Søgaard, Anders; Maaløe, Lars; Igel, Christian.
Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Bind 2020-October International Speech Communication Association (ISCA), 2020. s. 4352-4356.Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Do end-to-end speech recognition models care about context?
AU - Borgholt, Lasse
AU - Havtorn, Jakob D.
AU - Agic, Željko
AU - Søgaard, Anders
AU - Maaløe, Lars
AU - Igel, Christian
PY - 2020
Y1 - 2020
N2 - The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual information in the audio input. We find that the AED model is indeed more context sensitive, but that the gap can be closed by adding self-attention to the CTC model. Furthermore, the two models perform similarly when contextual information is constrained. Finally, in contrast to previous research, our results show that the CTC model is highly competitive on WSJ and LibriSpeech without the help of an external language model.
AB - The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual information in the audio input. We find that the AED model is indeed more context sensitive, but that the gap can be closed by adding self-attention to the CTC model. Furthermore, the two models perform similarly when contextual information is constrained. Finally, in contrast to previous research, our results show that the CTC model is highly competitive on WSJ and LibriSpeech without the help of an external language model.
KW - Attention-based encoder-decoder
KW - Automatic speech recognition
KW - Connectionist temporal classification
KW - End-to-end speech recognition
UR - http://www.scopus.com/inward/record.url?scp=85098151098&partnerID=8YFLogxK
U2 - 10.21437/Interspeech.2020-1750
DO - 10.21437/Interspeech.2020-1750
M3 - Article in proceedings
AN - SCOPUS:85098151098
VL - 2020-October
SP - 4352
EP - 4356
BT - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
PB - International Speech Communication Association (ISCA)
T2 - 21st Annual Conference of the International Speech Communication Association, INTERSPEECH 2020
Y2 - 25 October 2020 through 29 October 2020
ER -
ID: 254726027