English MFA ivector extractor v2.1.0#

Maintainer: Montreal Forced Aligner
Language: English
Model type: Ivector extractor
Features: MFCC
Architecture: ivector
Model version: v2.1.0
Trained date: 2023-01-04
Compatible MFA version: v2.1.0
License: CC BY 4.0
Citation:

@techreport{mfa_english_mfa_ivector_2023,
	author={McAuliffe, Michael and Sonderegger, Morgan},
	title={English MFA ivector extractor v2.1.0},
	address={\url{https://mfa-models.readthedocs.io/ivector/English/English MFA ivector extractor v2_1_0.html}},
	year={2023},
	month={Jan},
}

If you have comments or questions about this model, you can check previous MFA model discussion posts or create a new one.

Training corpora

Common Voice English v8.0:
- Hours: 2479.95
- Speakers: 74,811
- Utterances: 1,781,717
LibriSpeech English:
- Hours: 982.10
- Speakers: 2,484
- Utterances: 292,367
Corpus of Regional African American Language v2021.07:
- Hours: 124.31
- Speakers: 193
- Utterances: 236,792
Google Nigerian English:
- Hours: 5.77
- Speakers: 31
- Utterances: 3,359
Google UK and Ireland English:
- Hours: 31.29
- Speakers: 120
- Utterances: 17,877
NCHLT English:
- Hours: 56.43
- Speakers: 210
- Utterances: 77,412
ARU English corpus:
- Hours: 7.13
- Speakers: 12
- Utterances: 8,640

Installation#

Install from the MFA command line:

mfa model download ivector english_mfa

Or download from the release page.

Intended use#

This model is intended for speaker diarization and clustering.

Performance Factors#

As ivector extractors are trained on a large amount of data and without reference to language-specific resources, they may be useful outside of the specific language variety being trained, however, differences in the languages may impact performance. Particularly sociolinguistic aspects related to identity presentation, the linguistic status voice quality (breathy, creaky modal), and non-linguistic factors like recording conditions and microphone response may affect diarization performance across languages.

Metrics#

Speaker diarization systems are evaluated through Equal Error Rate (EER), the error when false accept rate is equal to the false rejection rate.

EER: 0%

Ethical considerations#

Deploying any Speech-to-Text model into any production setting has ethical implications. You should consider these implications before use.

Demographic Bias#

You should assume every machine learning model has demographic bias unless proven otherwise. For ivector extraction models, diarization of male speakers generally has better performance than diarization of female speakers, even with equal amounts of training data.

Surveillance#

Speech-to-Text technologies may be misused to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in many countries. You should not assume consent to record and analyze private speech.