Accuracy numbers

[MattWiethoff]

Hi there! Great project!

I was wondering what eval set you're using to get your accuracy numbers? I found them on your website: https://zamia.org/asr/

Also, any chance you know how it does on a conversational eval set (something like HUB5'00)?

Thanks!

[gooofy]

Hi Matt,

for each corpus you'll find a file called 'spk_tst.txt' which contains the name of the speakers whose recordings form the evaluation set.

cheers,

guenter