[COGA] 2.4 Speech Recognition and Speech Production

[RealJoshue108]

[Added on behalf of COGA TF]

• User Need 12: Consider expanding to include examples of atypical speech, such as:
  • Describing a word rather than using the word, which is common in people with age-related memory loss, aphasia/language disorder, and dyslexia
  • Using a similar word that is technically correct, but contextually not correct
  • Stutter, echolalia, and challenges with word-finding
  • Long pauses that may be caused by either discomfort speaking, word-finding challenges, or different processing speed capabilities
  • Slurred speech, such as that being worked on by Google’s Project Euphonia in the context of ALS
• REQ 12b: Consider adding a sentence that mentions artificial intelligence (AI) might be a tool to determine contextual accuracy and closeness to try to help with meaning, especially with “real word errors” (errors that are technically correct, but given the larger context and purpose, are wrong).
• REQ 13b: Suggest rewording, as noted here in brackets: “If the system's confidence in its recognition of the user's utterance is below a reasonable threshold, prompt the user to [remove: repeat or ] confirm [remove: the request made or the information spoken] in a way that supports working memory and reduces speaking effort. For example, the system could say, “I heard…Is that correct?”
  • COGA TF is suggesting these changes because for some individuals with speech differences, it takes a lot of effort to talk, asking the user to repeat could be a big ask.
• REQ 13c: Consider expanding to include examples of types of sounds other than speech, as noted here in brackets: “Allow the user to decide at any time to interact with the system via means of input other than speech, [add: such as grunts or claps,] even if a spoken dialogue is already in progress.”
• New note at the end of this section: Consider adding a note similar to the one in the previous section about sign language: “Some of these REQs are more challenging than others and are not yet feasible but are nevertheless stated here to encourage further research and development efforts. For now, all developers should ensure there are alternative ways to provide input in these cases.”

[RealJoshue108]

Mostly editorial with some additional suggested requirements

[RealJoshue108]

Discussed by Research Questions TF https://www.w3.org/2022/02/02-rqtf-minutes.html