- 
                Notifications
    You must be signed in to change notification settings 
- Fork 1
Open
Labels
Description
This question alone could be a PhD thesis. For purposes of this investigation, we're looking for some okay approximations depending on context.
Context: conformance to standards
Example of this context: where you can't check the source code
Working assumption: announced phonemes match the language's pronunciation rules and dictionary pronuncation keys, except for homographs and rare words
(How rare?)
Context: pragmatic accessibility
This is a usability question. I predict the answer will be highly variable depending on many factors such as:
- frequency and type of linguistic errors in the speech output
- user's language skills
- user's screen reader experience in the particular combination of language and synthesizer voice
- speech rate
- hearing acuity and audio environment
- nature of the intellectual or interactive task
There's probably some existing research on this.
I'll also ask a couple of users.