Skip to content

Latest commit

 

History

History
79 lines (62 loc) · 4.07 KB

README.md

File metadata and controls

79 lines (62 loc) · 4.07 KB
Raw

ABX-accent

The ABX-accent, project based on the preparation and the evaluation of Accented English Speech Recognition Challenge (AESRC) dataset[1], using ABXpy for the evaluation [2][3]. We provide in our GitHub repository all the necessary elements for the preparation and evaluation process.

The ABXpy metric evaluate for a pair of sounds representations (A,B) from for example (”bap”,”bop”), the probability that the representation X of another instance of the sound ”bap” is closer to A than B. ABX error rate is computed by averaging over all the minimal pairs of phone trigram in the corpus. In this benchmark, we focus on the harder ABX across speaker metric, which uses X instances from a different speaker than the one of the pair(A,B).

The Accented English Speech Recognition Challenge consists of ten different regional accents: American, British, Canadian, Chinese, Indian, Japanese, Korean, Portuguese, Spanish, Russian.

Start the project

All you need to get started to work on the development data and the evaluation of AESRC is :

The setup procedure is described for Linux. It has been tested on several distributions (Ubuntu 16.04, Debian Jessie and CentOS 6). It should work as well on macOS.

Organization

abx-accent/
├──  scripts
│   └── prepare/
│   │   └── data_splits
│   │   └── abkhazia
│   └── evals/
│   │   └── generate_item_files
│   │   └── generate_abx_score
│   │   └── abx_score_average
│   └── README.rst
├── data
│   └── prepare/
│   │   └── data_splits
│   │   └── abkhazia
│   │   │   └── forced_alignment
│   └── evals/
│   │   └── item_files
│   │   │   └── dev_set
│   │   │   └── test_set
│   │   └── abx_score
│   │   │   └── across_task
│   │   │   └── within_task
│   │   └── abx_score_average
│   │   │   └── dev_set
│   │   │   └── test_set
│   └── README.rst
│README.rst

License

Copyright 2022 CoML team (Inria, ENS, CNRS, EHESS)

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

References

  • [1] Xian Shi, Fan Yu, Yizhou Lu, Yuhao Liang, Qiangze Feng, Daliang Wang, Yanmin Qian, and Lei Xie, “The accented english speech recognition challenge 2020: open datasets, tracks, baselines, results and methods,” in ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).IEEE, 2021, pp. 6918–6922.

  • [2] Ewan Dunbar, Julien Karadayi, Mathieu Bernard, Xuan-Nga Cao, Robin Algayres, Lucas Ondel, Laurent Besacier, Sakriani Sakti, and Emmanuel Dupoux, “The zero resource speech challenge 2020: Discovering discrete subword and word units,” 2020.

  • [3] Ewan Dunbar, Mathieu Bernard, Nicolas Hamilakis, Tu Anh Nguyen, Maureen de Seyssel, Patricia Roz ́e, Morgane Rivi`ere, Eugene Kharitonov, and Emmanuel Dupoux, “The zero resource speech challenge 2021: Spoken language modelling,” 2021.

  • paper?