- From: Nick Levinson <nick_levinson@yahoo.com>
- Date: Mon, 5 Nov 2018 22:48:16 +0000 (UTC)
- To: "www-voice@w3.org" <www-voice@w3.org>
- Message-ID: <652936216.1363964.1541458096165@mail.yahoo.com>
TTS could stress a word or string differently according to its position in a sentence (or comparable linguistic unit larger than a word, referred to here as a sentence). If TTS does so in general, but in one sentence the intrasentence stress is omitted, then the listener will understand it as not a complete sentence. IPA in a phoneme element deprives TTS of information about stress outside of a word. I suggest adding in the Pronunciation Lexicon Specification an attribute to the phoneme element, such as "position", that would have three possible values: "start", "end" and "startend". The value "start" would mean that the string within the phoneme element is the start of a sentence or similarly deserving linguistic unit larger than the string. The value "end" would mean that the string within the phoneme element is the end of a sentence or similarly deserving linguistic unit larger than the string. The value "startend" would mean that the string within the phoneme element is, for purposes of PLS and TTS, a complete sentence so that stresses are to be managed accordingly. PLS would make no grammatical judgment about what is a sentence; for example, sentence fragments and run-on sentences would be treated by PLS as proper sentences for both PLS and TTS. TTS could make its own judgments about anything, as far as PLS is concerned. -- Nick
Received on Monday, 5 November 2018 22:48:41 UTC