Click here to register.


Way to get more data
User: pedro_loures
Date: 6/25/2014 10:41 pm
Views: 4464
Rating: 21

I've seen that Voxforge already have a way to use Librivox to help, but why isn't you guys taking advantage of other languages in that as well? There are many books from Brazil in there, and they would be great helping to add quite a lot to the small database Voxforge have so far in this language.


Eu vi que vocês já tem um jeito de usar o Librivox para ajudar, mas por que vocês não tiram vantagem de outras línguas também? Tem vários livros brasileiros lá, e eles seriam muito úteis em adicionar um tanto nessa pequena base de dados que o Voxforge tem em português até agora.

Re: Way to get more data
User: nsh
Date: 6/26/2014 4:31 pm
Views: 164
Rating: 21

Hello Pedro

This is a great idea.

Unfortunately before training you need to segment book on sentences. Work on automatic segmentation is being done but it is not complete.

It would be helpful if you could collect books in some regular format for the training - audio and text files. Then automatic tool could train models from them.


Re: Way to get more data
User: Visitor
Date: 6/26/2014 7:45 pm
Views: 2066
Rating: 18

I have some audios and I'm finding the correct text, because I have to listen and check every chapter of the book, after all, there may be a word different, for example. Would be great if they put the web page where they used to read it, but, what to do, hahaha. When I get it done, I'll post in this topic the links :)