search

laurensw75 / kaldi_egs_CGN

Kaldi recipe for creating Dutch ASR from CGN
7 stars 5 forks source link

NBest-corpus #1

Closed psmit closed 6 years ago

psmit commented 7 years ago

Hi,

I have been looking to obtain the Nbest corpus to evaluate my own CGN-recipe. I notice in your dataprep that your development set is created using a file named "nbest". Is the Nbest development set a subset of CGN?

Do you know a place I could obtain the evaluation part of that corpus? TST-Centrale does offer it anymore...

Regards,

Peter Smit (peter.smit@aalto.fi)

laurensw75 commented 7 years ago

Hi Peter,

When we participated in the NBest benchmark, the development set was indeed a part of CGN. As it's been a while, I cannot be 100% sure which parts of CGN were in fact development set, but the list I used are the most likely ones. I am also not sure where to get the NBest corpus, but perhaps David van Leeuwen d.vanleeuwen@let.ru.nl can tell you more about this. Also, I think Spex was involved, so Henk van den Heuvel h.vandenheuvel@let.ru.nl would be another option. If they have no objections to me sharing it, if needed, I would be happy to provide you with what I have.

In the mean time, I am really interested in your CGN-recipe :-)

Best,

Laurens

On Thu, Aug 31, 2017 at 6:40 AM, Peter Smit notifications@github.com wrote:

Hi,

I have been looking to obtain the Nbest corpus to evaluate my own CGN-recipe. I notice in your dataprep that your development set is created using a file named "nbest". Is the Nbest development set a subset of CGN?

Do you know a place I could obtain the evaluation part of that corpus? TST-Centrale does offer it anymore...

Regards,

Peter Smit (peter.smit@aalto.fi)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/laurensw75/kaldi_egs_CGN/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/APQTvrqOt_TZLdORA0jCtT_722L3uNffks5sdjkhgaJpZM4PISw3 .