Keywords: speech recognition, re-speaking
Abstract: This paper presents two user studies that investigate how errors that occur during speech recognition affect users’ text entry performance and experience. For our work, we used a speech recognition system that injects believable errors in a controlled manner, and where users could fix errors by re-speaking a small part of their original utterance. Participants were asked to transcribe a set of phrases using our system, either with or without the insertion of errors, In the first study, we injected up to 33% errors, but saw no substantial results. Yet, participants commented consistently on the used phrase set, which did not correspond well with spoken English.
Thus, we created a novel phrase set based on spoken phrases.
In our second study, we used this phrase set and inserted errors into the speech recognition results with either 25% or 50% probability. The results showed that inserting errors in the speech recognition system had a significant effect on participants' perceived mental workload. In addition, we find that inserting errors increased the number of errors users made during the task.
According to our findings, users have a fairly high tolerance for errors encountered in speech transcription.
Track: HCI/visualization
Revision: No
Revision Reviewers: No opinion
3 Replies
Loading