Re: Forvo - Useful for acoustic modeling?
Peter Grasch <[email protected]>
| Newsgroups | gmane.comp.kde.devel.accessibility |
|---|---|
| Message-ID | <[email protected]> |
On 08/03/2013 02:46 PM, Simon Eichhammer wrote: > I just scanned my bookmarks and found "Forvo". > [...] > This could be useful for acoustic modeling. In my opinion: No, these recordings are not particularly useful for acoustic model training as such. The recognition situation needs to mirror the training data as closely as possible - everything else is largely contra productive. They may be useful for dialect classification later on - something we're not doing yet (and shouldn't be concerning ourselves with at this point). Best regards, Peter