Shopper speech recognition programs don’t work as effectively for many individuals with speech variations, corresponding to stuttering, relative to the remainder of the overall inhabitants. Nonetheless, what will not be clear is the diploma to which these programs don’t work, how they are often improved, or how a lot folks wish to use them. On this paper, we first handle these questions utilizing outcomes from a 61-person survey from individuals who stutter and discover contributors wish to use speech recognition however are ceaselessly minimize off, misunderstood, or speech predictions don’t symbolize intent. In a second research, the place 91 individuals who stutter recorded voice assistant instructions and dictation, we quantify how dysfluencies impede efficiency in a consumer-grade speech recognition system. By means of three technical investigations, we show what number of frequent errors could be prevented, leading to a system that cuts utterances off 79.1% much less typically and improves phrase error fee from 25.4% to 9.9%.
