Releases: coqui-ai/STT-models
Dhivehi STT v0.1.1
Dhivehi STT v0.1.1 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Dhivehi / ދިވެހި /
dv - Model date: April 26, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.1 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{dhivehi-stt, author = {Tyers,Francis}, title = {Dhivehi STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-DV-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Dhivehi Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 91.2% | 29.3% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Chuvash STT v0.1.1
Chuvash STT v0.1.1 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Chuvash / Чӑвашла /
cv - Model date: April 26, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.1 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{chuvash-stt, author = {Tyers,Francis}, title = {Chuvash STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-CV-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Chuvash Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 95.4% | 33.7% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Breton STT v0.1.1
Breton STT v0.1.1 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Breton / Brezhoneg /
br - Model date: April 26, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.1 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{breton-stt, author = {Tyers,Francis}, title = {Breton STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-BR-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Breton Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 89.1% | 37.7% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Italian STT 2020.8.7
Italian STT 2020.8.7 (Mozilla Italia)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained and released by the Mozilla Italia Community.
- Model language: Italian / italiano /
it - Model date: August 7, 2020
- Model type:
Speech-to-Text - Model version:
2020.8.7 - Compatible with 🐸 STT version:
v0.9.3 - License: CC0
- Citation details:
@techreport{italian-stt, author = {Mozilla Italia}, title = {Italian STT 2020.8.7}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-IT-2020.8.7} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Italian Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported by Mozilla Italia (the transfer learning model).
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 28.7% | 11.8% |
| M-AILABS | 15.0% | 4.2% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: 1.0
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on:
- ~130 hours of the Common Voice Italian dataset
- ~127 hours of the M-AILABS Italian dataset
Evaluation data
The Model was evaluated on Common Voice and M-AILABS.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Slovenian STT v0.1.0
Slovenian STT v0.1.0 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Slovenian / Slovenščina /
sl - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.0 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{slovenian-stt, author = {Tyers,Francis}, title = {Slovenian STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-SL-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Slovenian Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 90.2% | 31.1% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Sakha STT v0.1.0
Sakha STT v0.1.0 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Sakha / Саха тыла /
sah - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.0 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{sakha-stt, author = {Tyers,Francis}, title = {Sakha STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-SAH-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Sakha Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 96.3% | 37.9% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Romanian STT v0.1.0
Romanian STT v0.1.0 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Romanian / Românește /
ro - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.0 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{romanian-stt, author = {Tyers,Francis}, title = {Romanian STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-RO-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Romanian Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 92.9% | 34.9% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Comodoro STT v0.1.0
Czech STT v0.1.0 (Vojtěch Drábek)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Vojtěch Drábek.
- Model language: Czech / čeština /
cs - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.0 - Compatible with 🐸 STT version:
v0.9.3 - License: CC-BY-NC
- Citation details:
@techreport{chuvash-stt, author = {Drábek,Vojtěch}, title = {Czech STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CS-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Czech Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
More information reported on Github.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 44.6% | 11.2% |
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on the following corpora:
- Vystadial 2016 – Czech data
- OVM – Otázky Václava Moravce
- Czech Parliament Meetings
- Large Corpus of Czech Parliament Plenary Hearings
- Common Voice Czech
- Some private recordings and parts of audioboooks
Evaluation data
The model was evaluated on Common Voice Czech.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
English STT v0.9.3
English STT v0.9.3 (Coqui)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Maintained by Coqui.
- Model language: English / English /
en - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.9.3 - Compatible with 🐸 STT version:
v0.9.3 - License: MPL
- Citation details:
@techreport{english-stt, author = {Coqui}, title = {English STT 0.9.3}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-EN-0.9.3} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the English Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
More detail on model training and evaluation can be found in the release notes.
Real-Time Factor
Real-Time Factor (RTF) is defined as processing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: 0.66
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on the following corpora: Fisher, LibriSpeech, Switchboard, Common Voice English, and 1,700 hours of transcribed NPR (WAMU) radio shows explicitly licensed to use as training corpora.
Evaluation data
The Model was evaluated on the LibriSpeech clean dev corpus as validation data, and LibriSpeech clean test as testing data.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.
Portuguese STT v0.1.0
Portuguese STT v0.1.0 (ITML)
Jump to section:
- Model details
- Intended use
- Performance Factors
- Metrics
- Training data
- Evaluation data
- Ethical considerations
- Caveats and recommendations
Model details
- Person or organization developing model: Originally trained by Francis Tyers and the Inclusive Technology for Marginalised Languages group.
- Model language: Portuguese / Português /
pt - Model date: April 9, 2021
- Model type:
Speech-to-Text - Model version:
v0.1.0 - Compatible with 🐸 STT version:
v0.9.3 - License: AGPL
- Citation details:
@techreport{portuguese-stt, author = {Tyers,Francis}, title = {Portuguese STT 0.1}, institution = {Coqui}, address = {\url{https://github.com/coqui-ai/STT-models}} year = {2021}, month = {April}, number = {STT-CV6.1-PT-0.1} } - Where to send questions or comments about the model: You can leave an issue on
STT-modelissues, open a new discussion onSTT-modeldiscussions, or chat with us on Gitter.
Intended use
Speech-to-Text for the Portuguese Language on 16kHz, mono-channel audio.
Performance Factors
Factors relevant to Speech-to-Text performance include but are not limited to speaker demographics, recording quality, and background noise. Read more about STT performance factors here.
Metrics
STT models are usually evaluated in terms of their transcription accuracy, deployment Real-Time Factor, and model size on disk.
Transcription Accuracy
The following Word Error Rates and Character Error Rates are reported on omnilingo.
| Test Corpus | WER | CER |
|---|---|---|
| Common Voice | 84.1% | 32.5% |
Real-Time Factor
Real-Time Factor (RTF) is defined as proccesing-time / length-of-audio. The exact real-time factor of an STT model will depend on the hardware setup, so you may experience a different RTF.
Recorded average RTF on laptop CPU: ``
Model Size
model.pbmm: 181M
model.tflite: 46M
Approaches to uncertainty and variability
Confidence scores and multiple paths from the decoding beam can be used to measure model uncertainty and provide multiple, variable transcripts for any processed audio.
Training data
This model was trained on Common Voice 6.1 train.
Evaluation data
The Model was evaluated on Common Voice 6.1 test.
Ethical considerations
Deploying a 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 STT models, it is often the case that transcription accuracy is better for men than it is for women. If you are using this model in production, you should acknowledge this as a potential issue.
Surveillance
Speech-to-Text may be mis-used to invade the privacy of others by recording and mining information from private conversations. This kind of individual privacy is protected by law in may countries. You should not assume consent to record and analyze private speech.
Caveats and recommendations
Machine learning models (like this STT model) perform best on data that is similar to the data on which they were trained. Read about what to expect from an STT model with regard to your data here.
In most applications, it is recommended that you train your own language model to improve transcription accuracy on your speech data.