Result Details
The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units
KARADAYI, J.
BERNARD, M.
CAO, X.
ALGAYRES, R.
ONDEL YANG, L.
BESACIER, L.
SAKTI, S.
Dupoux Emmanuel, FIT (FIT)
We present the Zero Resource Speech Challenge 2020, whichaims at learning speech representations from raw audio signalswithout any labels. It combines the data sets and metrics fromtwo previous benchmarks (2017 and 2019) and features twotasks which tap into two levels of speech representation. Thefirst task is to discover low bit-rate subword representations thatoptimize the quality of speech synthesis; the second one is todiscover word-like units from unsegmented raw speech. Wepresent the results of the twenty submitted models and discussthe implications of the main findings for unsupervised speechlearning.
zero resource speech technology, speech synthesis,acoustic unit discovery, spoken term discovery, unsupervisedlearning
@inproceedings{BUT168147,
author="DUNBAR, E. and KARADAYI, J. and BERNARD, M. and CAO, X. and ALGAYRES, R. and ONDEL YANG, L. and BESACIER, L. and SAKTI, S. and DUPOUX, E.",
title="The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units",
booktitle="Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",
year="2020",
journal="Proceedings of Interspeech",
volume="2020",
number="10",
pages="4831--4835",
publisher="International Speech Communication Association",
address="Shanghai",
doi="10.21437/Interspeech.2020-2743",
issn="1990-9772",
url="https://www.isca-speech.org/archive/Interspeech_2020/pdfs/2743.pdf"
}Moderní metody zpracování, analýzy a zobrazování multimediálních a 3D dat, BUT, Vnitřní projekty VUT, FIT-S-20-6460, start: 2020-03-01, end: 2023-02-28, completed