DeepSpeech 0.9.1 (Ionawr / January 2021)
Dyma'r sgriptiau a ddefnyddiwyd i hyfforddi a gwerthuso modelau Cymraeg DeepSpeech Mozilla ym mis Ionawr 2021 (21.01) ac i gyd-fynd â DeepSpeech fersiwn 0.9.1
These are the scripts used to train the Mozilla DeepSpeech Welsh models in January 2021 (21.01) and to coincide with DeepSpeech 0.9.1
Mae'r sgriptiau yn defnyddio data CommonVoice Mozilla Cymraeg a chyhoeddwyd ym mis Rhagfyr 2020 sydd yn cynnwys 117 awr o recordiadau sain.
The scripts use Welsh language data from Mozilla CommonVoice that was published in December 2020 and contains 117 hours of speech recordings.
Er bod maint y data Cymraeg yn CommonVoice yn annigonol ar gyfer hyfforddi adnabod lleferydd Cymraeg cyflawn, mae'r sgriptiau yn llwyddo i ddarparu modelau defnyddiol ac ymarferol drwy ddefnyddio dulliau dysgu trosglwyddol i addasu modelau Saesneg DeepSpeech gyda data Cymraeg. Mewn arbrofion syml a chynhelir gan y sgriptiau, mae’r model trawsgrifio yn camddeall tua 11% o eiriau mewn brawddeg, tra bo'r modelau ar gyfer cynorthwyydd personol Macsen yn camddeall 13%. Yn anffodus rydym yn amau bod wir ganrannau camgymeriadau yn lawer uwch petai ddull a data profi gwell ar gael.
Although the amount of Welsh data in CommonVoice is insufficient for training a complete Welsh language speech recognition engine, these scripts still succeed in providing a practical capability by using transfer learning methods to adapt Mozilla DeepSpeech project's English models with Welsh data. In simple evaluations that can be repeated with these scripts, the models for transcribing exhibit a word error rate of 11%, while models for a Welsh language voice assistant called Macsen have an error rate of 13%. Unfortunately we suspect that the true error percentages are a lot higher if better test data and methods were available
Modelau / Models
Mae'r modelau sydd wedi eu hyfforddi gyda'r sgriptiau hyn ac ar gael isod, sef / The trained models with these scripts are available below are
- model acwstig / acoustic model: techiaith_bangor_21.01.pbmm
- model iaith (parth trawsgrifio) / language model (transcription domain): techiaith_bangor_transcribe_21.01.scorer
- model iaith (parth macsen) / language model (macsen/voice assistant domain): techiaith_bangor_macsen_21.01.scorer
Enghreifftiau Defnydd / Example Usage
Mae enghraifft o sut ddefnyddir y ffeiliau modelau o fewn cynnyrch feddalwedd yn ein projectau Trawsgrifiwr.
Our Trawsgrifiwr (Transcriber) provides an example of how the models can be used in a software project
Yn ogystal a'n project gwasanaeth API adnabod lleferydd ar-lein
As well as our online speech recognition API project
https://github.com/techiaith/docker-deepspeech-cy-server/blob/v20.06/server/wsgi.py
Mae rhagor o enghreifftiau ar gael o ddogfennaeth Mozilla / Further examples are available from the Mozilla website
Python: https://deepspeech.readthedocs.io/en/latest/Python-Examples.html
C/C++: https://deepspeech.readthedocs.io/en/latest/C-Examples.html
Diolch / Thank you
Rydym yn ddiolchgar iawn i Rhoslyn Prys (meddal.com) a ymgymerodd â nifer o ymgyrchoedd torfoli ar sail wirfoddol, i'r Mentrau Iaith, Cyngor Gwynedd, Llyfrgell Genedlaethol Cymru a weithiodd gyda Rhoslyn ar rai o'r ymgyrchoedd hyn, hefyd i Lywodraeth Cymru, ac i liaws o gyfranogwyr ar draws Cymru a thu hwnt sydd wedi cyfrannu eu lleisiau i Common Voice Cymraeg.
We are very grateful to Rhoslyn Prys (meddal.com) who undertook on a voluntary basis a number of crowdsourcing campaigns, to the Mentrau Iaith, Gwynedd Council, the National Library of Wales who worked with Rhoslyn on some of these campaigns, to the Welsh Government, and to the many participants across Wales and beyond who have contributed their voices to the Welsh Common Voice datasets.