search

ccoreilly / LocalSTT

Android Speech Recognition Service using Vosk/Kaldi and Mozilla DeepSpeech
GNU General Public License v3.0
94 stars 27 forks source link

English models by default #2

Open kcpr opened 3 years ago

kcpr commented 3 years ago

Hi.

From what I've found out the models used by default are for Catala language. Have You considered adding English one as a default? I believe it would be useful for a great amount of people (including me). In my opinion there could even be two or even more separate APKs to download: for Catala and English language, depending on the models. That would be ok for me at least. Then You could even add this application to Google Play or F-Droid store I believe.

For now the application looks promising, but without English models it's not really useful for me. I may find the models and rebuild the application manually, but it's truly more effort then downloading an APK (and it's theoretically something that every English speaking person would have to do).

Thank You for Your work anyway. I don't feel there are many more free and open-source projects like that, if there are any. I appreciate it.

ftyers commented 3 years ago

It would be great to support more languages. I don't agree that the model should be English by default, but there should be a way to import models. I have trained models for ~31 compatible languages and they are available here.

kcpr commented 3 years ago

Thank You for the link and the opinion. Could You give any argument why English shouldn't be a default one though? It's an international language and this project may be useful for much greater amount of people out of the box, I believe that's fair enough, but maybe I'm wrong.

ftyers commented 3 years ago

Quan tu dius "llengua internacional", jo dic "llengua colonial". A mi no m'interessa aumentar el espaí on viu la llengua colonial, potser a tu sí que t'interessa. No dic que és millor o pijor, només que a mi no m'interessa. (Per cert si no entens el català la traducció automàtica pot donar-te un idea.)

kcpr commented 3 years ago

I could probably consider this sudden language change and suggestion to use a automatic translation impolite, but it's ok, we can do it this way. And thanks for the link, at first I just used Google Translate though.

Thank You for the answer. Still, the argument You presented seems to me like a quite personal opinion. I think I do not want to get into politics, but I can imagine that You may not be the fan of English. I also do not want to repeat myself, but can just present some statistics to confirm my thesis: https://www.statista.com/statistics/266808/the-most-spoken-languages-worldwide/ . And as I already suggested: I'd be totally ok with having two separate APKs or - what recently came to my mind - branches: for English and Catala.

And just to let You know: English isn't my native language.

By the way, for anyone interested in using English models: I wrote how I got them and how I used them in this issue: https://github.com/ccoreilly/LocalSTT/issues/3 . I haven't found any smaller DeepSpeech ".scorer" file though I think. I would be grateful if someone who found any could link it.

An automatic translation into Catala below (from the site You suggested): / Una traducció automàtica al català a continuació (des del lloc que heu suggerit):

Probablement podria considerar aquest sobtat canvi de llenguatge i el suggeriment d'utilitzar una traducció automàtica com a impolita, però està bé, podem fer-ho d'aquesta manera. I gràcies per l'enllaç, al principi vaig fer servir Google Tradueix.

Gràcies per la resposta. No obstant això, l'argument que ha presentat em sembla una opinió bastant personal. Crec que no vull entrar en la política, però puc imaginar que no ets el fan de l'anglès. Tampoc vull repetir-me, però només puc presentar algunes estadístiques per confirmar la meva tesi: https://www.statista.com/statistis/266808/the-most-spite-languages-worldwide/ . I com ja he suggerit: estaria totalment d'acord amb tenir dues branques separades o -el que m'ha semblat recentment-: per a l'anglès i el català.

I només per fer-vos saber: L'anglès no és la meva llengua materna.

Per cert, per a qualsevol interessat en l'ús de models anglesos: vaig escriure com els vaig obtenir i com els vaig utilitzar en aquest tema: https://github.com/ccoreilly/LocalsTT/issues/3 . No he trobat cap fitxer més petit de DeepSpeech ".scorer", però crec. Li agrairia que algú que hagi trobat alguna cosa la vinculés.