Minutes of today's Personalization TF call

Hello Personalization and Internationalization,

Thank you for attending our call today and for the positive discussion and suggestions as to our next steps. Please find the minutes at https://www.w3.org/2021/09/13-personalization-minutes.html (apologies for the odd formatting towards the end) and below. Looking forward to continuing to work with you.

Best regards,


Matthew

– DRAFT – Personalization Task Force Teleconference

13 September 2021

Attendees

Present
    addison, becky, CharlesL, janina, JF, Lionel_Wolberger, Matthew_Atkinson, mike_beganyi, Roy, sharon_, xfq

Regrets
    -

Chair
    -

Scribe
    Lionel, Matthew_Atkinson

Contents

1.  i18n issue #144

Meeting minutes

present!

<JF> Preseent+

present?

<Lionel_Wolberger> https://github.com/w3c/personalization-semantics/issues/144


i18n issue #144

Lionel_Wolberger: Thank you to i18n for joining

janina: AAC (Augmented and Alternative Communication) used by people with certain cognitive disabilities who use graphical symbols to communicate. Background well before the web. In the 60s/70s boards with symbols on were used to denote concepts.

janina: The symbols used depended on where you were taught. So multiple symbol sets came to be. The one that was most commecrially successful has created an index that's useful for translating between different symbol sets.

janina: BLISS has given us royalty-free access to their index of symbols. It's essentially a table.

janina: We are aiming to use this in our standard. Therefore, to convey the concept of "help" we would be able to use the symbol in the language/symbol set the user knows.

janina: We want to reach "the next billion users" on the web.

<Roy> Bliss symbol table

<Lionel_Wolberger> <span data-symbol="13621 12324 17511">cup of Tea</span>

Lionel_Wolberger: We think it was when there were multiple symbols that i18n got involved.

<JF> https://www.w3.org/WAI/APA/video-examples


JF: The video (at the above URL) shows what we're trying to accomplish. We are not so much translating between languages, but transforming from one symbol set to another.

JF: We aim to provide an approximation of the meaning from the text.

Lionel_Wolberger: Lots of symbol-based communicxations today (save button icon; emoticons/emoji).

Lionel_Wolberger: The markup we're proposing allows authors to specify and user agents to interpret the symbols. E.g. could provide a symbol to decorate an input field that's been marked up with our "purpose" attribute.

JF: What we're doing here is creating a taxonomy. Not concerned about the symbols themselves, but we are using the existing BLISS taxonomy to indicate concepts. The appropriate symbol from the user's chosen set can then be inserted.

addison: Thanks for summary; recall previous TPAC meetings. Issue 144 maybe has more than one thing in it. We have several concerns.

addison: Text direction (ltr/rtl). Symbols are not text characters; the rendering algorithm for text can't affect them. Will the symbols be rendered in the correct order? Is there a way to indicate the direction in which they should be ordered?

<JF> It inherits the direction from the source text

addison: When you transform a natural language text to an augmented representation, the language of the text might be important. Word order and symbol order across languages may vary.

addison: Therefore some concepts of grammar would be needed (consider a fill-in-the-blanks game: the placing of the symbols that represent the blanks depends on the grammar).

r12a: Not sure that the order would be inherited when images are inserted.

r12a: E.g. symbols for man hit boy as rendered might've meant boy hit man.

addison: There could be other issues. How do we attach to the grammar of the underlying language? So far all examples have been 1:1 word mappings, which won't always be the case.

Lionel_Wolberger: The overarching concern: grammar and bidi considerations could cause the symbols to render in the wrong order.

addison: Original issue (144) is "how do you indicate the direction and, if necessary, the language of the symbols?"

addison: Then: how do you render things directionally correctly, and how do we do the word association, if necessary.

<Zakim> r12a, you wanted to react to addison

r12a: BLISS doesn't just cover symbols replacing words, but has its own way of providing grammatical information (such as plurals). Question: how accurately can the content be read if we don't support things like this?

JF: I think we're talking past each other. We're not creating a translation tool. Our symbols are being overlaid on top of text (thus inheriting direction and intent from the author).

JF: We are augmenting the existing text (example in our video). Our text in that example is written in English. The symbols are placed at the start of each word, and are only applied to the most important concepts on the screen.

JF: We're trying to make text that wasn't written using symbols easier to understand by adding this extra information.

JF: We've talked about personalization services being performed by proxy servers/services. We are just augmenting the text on the screen with symbols.

<JF> video here Richard: https://www.w3.org/WAI/APA/video-examples


r12a: What I'm missing from the spec is a picture of what it'd look like with something more complex than e.g. "a cup of tea"

<JF> augments

r12a: Does the picture replace the words, or are the words still there somewehre?

janina: We're trying to crawl before we walk/run. At some point the concerns that you have will apply, but we're not aiming to go that far at this point. We're trying to put semantic landmarks/waypoints into the page, for users who would otherwise get lost/confused.

<becky> images are here: https://github.com/w3c/personalization-semantics/wiki/Explainer-for-Personalization-Semantics#severe-language-impairment


janina: Another way we're doing this is by simplification (removing unnecessary content).

janina: We're going to be looking for/have implementations such as extensions to browsers that will be configured by someone assisting the user and will perform these translations after that point.

janina: It would be good if we could add an example that's not English/ltr.

<Roy> A chrome extension example

janina: We're relying on existing tech. There are people who are already using symbolic comms in similar ways and we're relying on what's typically done there for directionality.

janina: We should add another video demonstration in a rtl language.

janina: We're trying to make comms accessible to people who are not able to hae personal support.

<becky> there are some examples without pictures but that include conjugation at: https://github.com/w3c/personalization-semantics/wiki/Explainer-for-Personalization-Semantics#severe-language-impairment


<becky> corrected: there are some examples without pictures but that include conjugation at: https://w3c.github.io/personalization-semantics/content/#symbol-explanation


addison: ACK; undersood. We want to help you produce the right outcome. Looking at directionality first: does the order of the symbols matter, and does it match the order of the text? May also have some other-language words intermixed. Where are the symbols placed?

janina: These questions would probably be of interest to screen reader users too.

addison: We've ensured text has those kinds of markers, but if inserting symbols, does there need to be control over how the symbols are placed on the screen that's provided to the author. If so, is it just the direction of the text or is it something else?

JF: It's inherited from the source, as we're not doing translation here.

<CharlesL> r

JF: At the current time, things like direction are not covered by the spec. Currently we just literally convey what the author wrote.

janina: Action proposed above; we should find out what the common practice is.

<CharlesL> Example "Dog bites man" is defined by <span data-symbol="1">Dog</span><span data-symbol="2">bites</span><span data="man">man</span>

Lionel_Wolberger: "man bites dog" example from JF: is each word annotated with its symbol, or is the phrase as a whole annotated? I think there would be no issue if it was words rather than phrases.

<Zakim> JF, you wanted to also note that we don't have any specific tooling at this time

JF: The spec as-is allows for both. We can make a phrase, with space-separated values. Or word-by-word direct augmentation (important word).

Lionel_Wolberger: If we said in this version that there must be only one symbol at most per word, would that solve the issue?

janina: I think we should find out what is standard practice within educational environments currently (the tech predates personal computing).

<r12a> i think you need to test this

<CharlesL> Example "Dog bites man" is defined by <span data-symbol="1">Dog</span><span data-symbol="2">bites</span><span data-symbol="3">man</span>

CharlesL: Was thinking that at first each word would have its own symbol.

CharlesL: Then the whole thing would be rendered appropriately by the UA.

CharlesL: Would the symbol be placed correctly?

CharlesL: For some phrases, there may be a single symbol that encapsulate the phrase (or not).

CharlesL: We should look into ruby styling/placement.

<Lionel_Wolberger> Lionel asks Matthew: shall I scribe for you so you can contribute?

r12a: If you had that example but the words remain then that would probably solve the problem due to the bidi algorithm. If the words weren't there I think this wouldn't work and thus would be problematic.

r12a: What we're asking for is for you to show us how it works. Agree with JF but we need to see how it will work.

addison: Wanted to call out the similarity to ruby markup. Sometimes you mark up individual words, or groups of words/phrases.

addison: We're not here to hold you back; want to help you think about these issues and whether additional structure is needed to get the right results. +1 to r12a's request.

addison: We think there could be issues wrt rendering etc. but would need to know more.

<Lionel_Wolberger> Matthew_Atkinson: The tea (WikiHow) example, the symbols appear at the beginnings of words. Would the symbol have to presented in a left or right position?

<Matthew_Atkinson> Matthew_Atkinson: Maybe covered by the talk on individual word mark-up/ruby above... in the cup of tea example, the symbols are placed before the word. Would this work in rtl?

<Lionel_Wolberger> r12a: In the WikiHow example, the images come before the word in LTR, so the word would come before the word in RTL

<Lionel_Wolberger> ... there is one case where there are two symbols. that is where the direction of the symbols comes into play.

<Matthew_Atkinson> r12a: If you have single symbols per word, and the words included, this would work as it should. If you have multiple symbols, that may remain ltr within the rtl flow. Also if there's any directional bias in the images themselves, that will not be catered for (may/may not be a problem).

<Lionel_Wolberger> ... more than one symbol side by side is the issue

<Matthew_Atkinson> r12a: Important to test (as above).

<Matthew_Atkinson> janina: Proposed actions seem to make sense: we're gonig to be asked these questions in future. Need to be prepared. Expand on the demo; looking into ruby; relevant to ePub.

<xfq> +1

<CharlesL> +1 Janina, agree on the Publishing use cases

<Matthew_Atkinson> janina: We want to make sure the mechansims are present in the HTML stack. Needs to work in different contexts such as ePub in particular (as this may be a focus for us initally).

<Matthew_Atkinson> janina: We should revisit this during TPAC? Maybe already resolved 144 by then, but matters to the wider community so we should be ready.

<CharlesL> +1 to TPAC discussion with Publishing on this issue and ruby similarities.

<r12a> btw, please don't assume you can use ruby markup for this ! there may be similarities if you're annotating stuff, but ruby is semantically CJKM specific markup

<Matthew_Atkinson> Lionel_Wolberger: How about the first para of A Tale of Two Cities; would that be sufficient?

<Matthew_Atkinson> r12a: Yes

<Matthew_Atkinson> janina and becky: That would be good eventually, but it's not where we are now.

<Matthew_Atkinson> Lionel_Wolberger: But the wikiHow example is too simple?

<Matthew_Atkinson> (general agreement)

<Matthew_Atkinson> Lionel_Wolberger: Let's try to settle on an example whilst we have i18n.

<Lionel_Wolberger> https://www.w3.org/TR/ruby-use-cases/


<Matthew_Atkinson> JF: ruby is markup that helps people pronounce words. There's nothing in the ruby spec that enforces quality. Direction/grammar are QA concerns; out of scope for us now, as based on the person who's dooing the mark-up (we assume they'll have some understanding).

<Zakim> JF, you wanted to ask addison about quality control with ruby

<addison> +1 to richard's point

<Matthew_Atkinson> r12a: Don't assume that you can use ruby markup for this; it's a conceptual similarity only. There's a paragraph below the heading in the wikiHow example that should be sufficiently complex for an example. Really important to think ahead to something like this.

<Matthew_Atkinson> addison: +1 to r12a. Main concern is that you consider the challenges, and demonstrate them in some way or clarify that they are out of scope for now. We would be happy to look at more examples.

<Matthew_Atkinson> janina, becky: wikiHow example para would be OK, as would first para of A Tale of Two Cities.

<Matthew_Atkinson> Lionel_Wolberger: How should we continue to collaborate?

<Matthew_Atkinson> r12a: GitHub over email or other things for me personally. Allows centralised discussion with history.

<addison> thanks personalization folks!

-- 
Matthew Tylee Atkinson
--
Senior Accessibility Engineer
TPG Interactive
https://www.tpgi.com

A Vispero Company
https://www.vispero.com

--
This message is intended to be confidential and may be legally privileged. It is intended solely for the addressee. If you are not the intended recipient, please delete this message from your system and notify us immediately.
Any disclosure, copying, distribution or action taken or omitted to be taken by an unintended recipient in reliance on this message is prohibited and may be unlawful.

Received on Monday, 13 September 2021 16:00:44 UTC