I feel that the category specific speech recognition will help accuracy compared with a general service like Siri. On a separate note, how does one go about cr

I feel that the category specific speech recognition will help accuracy compared with a general service like Siri. On a separate note, how does one go about creating a speech recognition backend like this? A point in direction would be helpful. Thanks!