If you happen to know ahead of time the exact bounds of the grammar being recognized, then you can substantially improve the STT quality. So if, for a given game-state, I know ALL possible legal commands, I can do that and give a high quality of service.
As a fallback, once the STT service parses some audio, it can give multiple results, each with a certain level of confidence. If I can "try them out" against the engine without actually executing the command, then I have the chance to further filter the list by valid and invalid options. Once more, increasing the quality of service.
jaynabonne wrote:Not of Quest, no. But there may be another IF engine written in Java.
jjaquinta wrote:I'm trying to gauge the difficulty in font-ending a question with a speech to text service.
If you happen to know ahead of time the exact bounds of the grammar being recognized, then you can substantially improve the STT quality. So if, for a given game-state, I know ALL possible legal commands, I can do that and give a high quality of service.
As a fallback, once the STT service parses some audio, it can give multiple results, each with a certain level of confidence. If I can "try them out" against the engine without actually executing the command, then I have the chance to further filter the list by valid and invalid options. Once more, increasing the quality of service.