Word recognition

[rablack97]

This may be a stupid question, but is there a way to train vox commando to learn words it can't recognize.

When calling up an artist with a strange name, you pretty much have to place it in the builder and have vox say it back to you so you know how to say it so the system can recognize it.

For example, artist Beyonce is pronounced "Be yaw" by the system, artist 2Pac is pronounced as 2 Pack.

Would you have to adjust the tags to read BEE Yon Say and 2 Pahck to have it recognize what you want w/o sounding stupid?

 

[Kalle]

This may be a stupid question, but is there a way to train vox commando to learn words it can't recognize.

When calling up an artist with a strange name, you pretty much have to place it in the builder and have vox say it back to you so you know how to say it so the system can recognize it.

For example, artist Beyonce is pronounced "Be yaw" by the system, artist 2Pac is pronounced as 2 Pack.

Would you have to adjust the tags to read BEE Yon Say and 2 Pahck to have it recognize what you want w/o sounding stupid?

 

Hi rablack97, yes there is an easy way to do this, for a example for XBMC - open in command editor the the "xbmcArtists.xml" with a double click on it. (You find it in the group "play music by name" - command "play artist")
Then click "open/edit" and search in the follow table the artist you wish to change. Just put a comma after the name and add another name, after this save the payload.
The only thing is, if you let VC generate a new XBMC payload for artists, this file is overwritten, so it is better you save this file with a other name, create a other "play artist" command like: "play special named artist xyz" and use your saved payload. If you use TTS for this command, VC speak both names, thats a bit crazy.

Kalle

PS: You can also edit the "original" name in the payload

[jitterjames]

the TTS engine will not necessarily say it the same way that the  recognition engine expects it to be pronounced, this is obvious when you notice that different TTS engines will pronounce things differently.

I think Kalle's idea of creating a custom payloadXML command for problem artists makes sense.  I haven't see the need personally.  The fact that it expects 2Pac to be pronounced as 2 pack makes perfect sense to me.  How would you pronounce it?  As for beyonce, one solution (not ideal I know) would be to actually change the spelling of her name in the mp3 metadata tags.

You may be able to train WSR to recognize certain words when dictating using WSR in word.  Do the windows built in speech recognition tutorials to see how it is done.