I am trying to improve recognition of users using very large grammar files, by trying to reprocess a user's audio input through more then one grammar file, it's a bit hard to explain but it's something like the following:
Say a user is asked for their favourite artist and song in the same sentence. The user says "Michael Jackson Thriller". I want to first load up a grammar that will just recognize an album name from thousands and it would look something like "<Wildcard> <Album>" and then depending on the album's name, it will load up another grammar which will have the songs in that grammar such as "<Song> <Wildcard>" all using the same input recorded from the user.
Basically once the user says the album for Thriller and song, a grxml file with Thriller's songs are placed in the second grammar file instead of every song by every album. So then instead of asking the user "What is the album?" and then afterwards "What is the song?", I do it all in one process.
Does anyone know if this is possible using VR Workflows or if they have done this before or know how?