13 November 2023


Alan, matatk, PaulG, Sam

Meeting minutes

Agenda Review, Membership & Announcements

Review Use Case

<PaulG> https://github.com/w3c/pronunciation/wiki/Reasons-for-Delays-in-Spoken-Content

PaulG: Two new sections at the bottom of the wiki page: thought groups; and teaching pronunciation.

PaulG: Thought groups has two subsections.

PaulG: *reads out the example* No heuristic is going to get this right.

PaulG: This is written correctly, yet is still ambiguous.
… The next one is about clarity. Introducing pauses between clauses to try to make the meaning clearer. This pausing is emphasized particularly when people are learning the language (they may not be inflecting, which would've provided other clues).
… There is no punctuation in these examples but there are natural pauses.
… This first one isn't as strong as the disambiguation example, but there are others below that add more.
… The pauses allow the listener time for processing.

Sam: I can see people pushing back on disambiguation, as it adds complexity to content markup.
… E.g. German is very precise in terms of grammar and punctuation. It does depend on the language.
… This reminds me of the challenge of dealing with baroque music (e.g. Bach). There isn't as much documentation around dynamics, compared to Beethoven, and romantic composers.
… There's very little instruction compared to things that are far more recent.
… So some people could push back on it, for content authors to be domain experts in specific languages. We don't understand the requirements for other languages than English.

PaulG: With music, the interpretation is part of the presentation. In these examples, the text could be part of a headline, for example. We are looking at examples where there _is_ intended meaing.

Sam: [raises example of the book "Eats, shoots and leaves"]

PaulG: We like those exmaples, but what we were talking about last week is that they're not as compelling as examples because when they have proper punctuation, the pauses will naturally appear, for disambiguation.
… I was striving for an example where there's something ambiguous being stated, and it has to be ambiguous because of pausing.

Sam: Are these edge cases?

PaulG: We're not saying that the obligation is always on the author, but we would like it to be an option.

PaulG: Music example: could be Bach jazz, or Bach swing, for example. We want to give authors the fine-grained control if they want to do it.
… Especially, e.g., when reading out instructions.
… Per last week we discussed structural things like lists (which simple heuristics can reliably help with) where there's no need for the author to use the fine-grained controls.
… For something like a news headline, or something that is fixed, but needs to be interpreted in a way that makes sense - that's where these examples are coming in.

Sam: Just playing Devil's advocate, to anticipate pushback, and articulate when these features are expected to be used. I like the examples.

PaulG: The next set of examples is about teaching punctuation.
… [refers to example in wiki page]
… These are instances where the meaning is entirely different if it's not punctuated (by the listener inserting pauses) correctly.
… This is the inverse of the first example (where the correct punctuation was provided).
… There may be more examples of this too.
… It occurred to me that these sorts of examples could be used by teachers to _teach_ punctuation/pronunciation.
… The problem is that the TTS is being used by the education provider, rather than e.g. recorded speech.
… TTS is definitely cheaper to implement for test providers.
… Maybe these examples not as strong (the use case is though).
… Was referring to how generalizable it is.
… I think we need to get to 1-3 really strong examples to include.

<PaulG> matatk: I can't think of any more suitable examples right now.

<PaulG> ...we have a fine line to tread between masking TTS bugs and author controlled pronunciation

<PaulG> we'll meet again 11/27

<Alan> https://www.thoughtco.com/syntactic-ambiguity-grammar-1692179

<PaulG> For next meeting, we'll try to collect, categorize, and rank our examples.

Action Items

Github Issues and examples

Other Business

Minutes manually created (not a transcript), formatted by scribe.perl version 221 (Fri Jul 21 14:01:30 2023 UTC).


No scribenick or scribe found. Guessed: matatk

All speakers: PaulG, Sam

Active on IRC: Alan, matatk, PaulG, Sam