Chemistry Best Practices for EPUB #

26 Mar 2019


janina, dauwhe, George, Avneesh, Barrett, franco, Bill_Kasdorf, CharlesL


<liam> I’m sorry, i have a conflict today; want to note: if SVG is chosen, the SVG should be annotated so the higher-level chemML (or whatever) can be reconstructed from it.

<scribe> scribenick: dauwhe

George: Any new folks want to introduce themselves
... any comments on minutes from last time?
... minutes approved
... lots of publishers here
... let's start with Volker
... any additions?

Volker: I sent something around yesterday
... in addition to what I've showed previously
... you can now tap through with tabindex and aria-describedby
... do people think this is sufficient?
... the speech is not fully set up yet
... this is javascript free

George: javascript free?

Volker: yes, the examples yesterday were just svg and ARIA

Dan: I was playing around with the molecules
... I could tell what you were doing, but I guess I don't have ARIA?

Volker: what screenreader?

Dan: Jaws

Volker: could you tap through SVG?

Dan: no

Volker: I hadn't tried JAWS

George: I didn't get the samples
... are there other approaches that people know about, other than what Volker is doing

<franco> https://progressiveaccess.com/chemistry/embedded.html

Avneesh: I was able to read everything with the down arrow. Don't know about tabbing
... it was depth first search
... if you know chem it's ok
... but might be confusing for a student

George: depth-first search, I've seen that with math, with equal sign at top

<Avneesh> https://progressiveaccess.com/chemistry/embedded.html

George: is chem set up like that?

Volker: the svg linearizes the hierarchical structure
... you get the top level, with functional groups and rings
... then you can dive in
... if you go depth-first, you get the top level, then the benzene, then the details of that...
... with the tab index I've tried to replicate breadth-first
... there are probably other ways, like with tree navigation in aria
... but i don't know if that works without JS

janina: I can't really talk about tree navigation :)

George: I have lots of thoughts
... first, I'm thinking of a young person in beginning chemistry
... it seems like there would need to be a set of training materials
... to get students started to learn how to... read.
... Charles, this seems like something the Diagram Center could create
... finding funding to do this properly would be essential
... this is not a spare-time project
... what do we think of putting together a sample epub
... at epubtest.org we have a test book for a11y
... we're going to redo the math book
... should we put together some different approaches for Chemistry and put it into a sample EPUB?

CharlesL: that could work
... there are major inconsistencies, as there is with the web

George: you want a web page first?

CharlesL: probably, to see the limitations
... I had some issues with voiceover
... and we've heard of issues with jaws
... we could bring this to the dev group in the diagram center, and see what needs to be added to the svg

Avneesh: the JS piece looks promising
... the drawback is that it won't work with all the epub reading systems which don't support JS
... the other plan was to remove JS, as Volker has done
... it will work with many reading systems, but it is confusing--not so easy for a student to read

CharlesL: I'd like to know from the publishers on the call
... do you have any books with JS?

George: that's the next agenda item--where are the publishers in how they do chemistry and use JS

Avneesh: and we need to know this for both MathML and Chemistry

Dan: we build our ebooks in inkling habitat
... I'm assuming there's JS
... then we push to our platform
... I don't quite know what occurs there
... I was thinking that these SVG images... our images are shared between print and digital
... I'm more concerned about creating a whole new set of images for digital
... I don't know we'd react to that

Bill_Kasdorf: are you saying the exact same file format and file?

Dan: one is CMYK, the file for digital is scaled differently
... it's the same image essentially
... if it meant accessible images, this is what I would like--the SVG

George: do you know if they're SVG?

Dan: they're raster/PNG

George: that's the starting point--most publishers have PNGs or the equivalent
... and are putting in alt text, which is not easy

Dan: and is a lot of money

George: with the alt text

Dan: yes
... this is building the a11y into the images themselves

George: volker, with math, mathjax generates alt text, a textual description

volker: yes

George: can this be done from a chemistry SVG?
... if we decorate the svg we could get a spoken or alt text automatically

volker: the whole text is automatically generated from the bitmap via image analysis
... so one could start with the PNG
... easier to start with MOLE file
... even if you only want alt text, you could generate that automatically and incorrectly

George: so this is like OCR

Volker: yes, it's sortof like OCR. Image analysis and recognition
... you could also start with other things, even wikipedia

Dan: I think that's super
... I want to see two things
... we can argue how we group these molecules
... but what's nice is that the student can explore the molecule
... I don't like alt text because it's giving you someone's interpretation of the molecule
... with this they can understand the molecule in their own way
... i hope we can standardize the descriptions

???: in assessment questions, I worry that this might give away answers

Volker: fully accessible diagrams can be explored with JS
... we've done a best practice for a dutch org
... with a chapter about exam questions
... and how we can generate descriptions, leaving out the parts that the students need to find for themselves
... you can still walk through them, but then you don't get some info

George: so you take the same SVG that you would use for normal reading, and modify it for test question?

Volker: yes, that's an option on the generation phase

???: we also use them in lecture power points which are not web-based. could we convert to another format?

Volker: most of these systems, like powerpoint, can use SVG
... you can always generate bitmaps again

Dan: that's a good point about assessment
... how do we just give the information that we expect them to use?

Steve Noble: with linear formulas, with a rules based approach to reading you could have an assessment rule instead of a learning rule

scribe: you could change so you hear H2O instead of water
... so for linear chem formula written in mathml you can use a rules-based approach to change the reading
... a similar approach could be used for molecular

Kerry: with regard to high-stakes assessment, we wouldn't want to see a molecule's rendering in a test
... I like the two modes, so students can get versed on how the conventions work
... so the assessment mode won't be a surprise

George: if we can standardize this, then students would be introduced to this in training materials

Volker: I'm sharing a link to this website
... where students learn how to use and navigate molecules, and ch8 has exercises

George: what are the next steps?
... it looks like we're starting with images, doing recognition, and generating SVG
... and Volker, is this a product you have, that's being sold?
... I don't know anyone else who has htis

Volker: the original idea was to sell as a product

Dan: this is something we would purchase, and do the conversions ourselves

Volker: yes
... we've also thought about something like mathjax, with sponsors and open source

Bill_Kasdorf: would a publishers license to that be usable by a vendor?

Dan: normally we contract out a lot of our work, and licensing is an issue
... it's a good point

CharlesL: or the conversion vendor gets a license

Dan: we went to a specialized alt text vendor; we'd have to find someone who would specialize in this

CharlesL: some of these textbooks would have end-of-chapter questions, so students would already see both detailed versions and assessment versions

Kerry: this is assuming that instructors use end-of-chapter questions

CharlesL: they're still exploring the molecules

George: we have 20 minutes left
... what do we do next?
... janina, you have the CG on domain knowledge
... is the knowledge domain group a higher level, or would the chemistry be developed there?
... what's your design?

janina: I can see that the purpose of that CG is to come up with the meta-structure
... but you could have content from multiple domains in the same publication
... but we're not focused on the details of a particular domain

Avneesh: it's about least common denominator for all domains

janina: how do we know what we have in a particular title?

George: here, we're starting from the bottom and moving up. the bottom is the chem solution
... and we push that to the knowledge domain, and see if an SVG approach works for other domains

janina: yes, we'd be looking for patterns we can replicate

George: sounds like we should march forward with this group, and get our best practices out
... can SVG be used inline?

dauwhe: yes

CharlesL: we didn't get a question of publishers using SVG, or how much JS they use

Franco: for our enhanced EPUBs we use JS
... we do use PNGs for chemical structures
... we're not really using SVG now

Bill_Kasdorf: since I bring up the vendor issue, vendors do what the publishers ask them to do
... vendors could do SVG

George: this inline issue... if I have some molecule inline, is it OK?
... that's pushing us towards a single technique, SVG, to solve the problem

CharlesL: is the question about inline SVG vs just using sup/superscript

Dan: for inline formulas, we would just put them in HTML
... that's how I'm going to read it out loud

janina: that's a problem for a text-to-speech engine
... that's another task... we need to identify the type of formula, to help determine how it's read

Steve Noble: now we're back to talking about linear chemical equations, which are commonly done using MathML

scribe: and then the speech rules apply whatever rule you select
... in the long run, we need to look at a typical textbook
... and look at all the kinds of information are included
... there's going to be a lot of inline chemical formula, lots of lewis things, the periodic table, lots of charts, etc
... this is similar to what we did with the diagram accessible image book
... take a book that's already out, and suggest how to do all the images
... that would be useful for the industry

CharlesL: I like an a11y chem book, and best practices
... I see three ways to do inline chem--html, svg, and mathml
... we need to come up with best practices

Volker: it's more important to tell people to annotate their preferred solution
... the annotation should be standardized, not necessarily the representation

George: with mathml you can't differentiate math from chem

Steve Noble: Mathplayer uses heuristices

scribe: you get 95% or something

Franco: can you elaborate on heuristics

Steve NOble: You'll parse what's there, and look for a pattern

scribe: if you see a capital H with subscript 2, and then a plus sign... it's likely a chem formula
... but there's also not enough information at times

George: is this guessing being done by AT, it will be all over the planet, and drive up the costs
... if we can find a more content-based solution

Volker: I have similar heuristics in mathjax
... most chem on the web is done with tex macros, and we have additional info available

Steve Noble: TTS has always used heuristics

scribe: if content authors told us, we wouldn't have to guess

Bill_Kasdorf: this is into what Janina is working on, a universal domain declaration

janina: yes, that's what I'm talking about
... and this might help stuff to work cross-platform
... and I think it's the content creators who will determine things, the people here

George: we have a couple of minutes to go
... I need help on the agenda for the next call
... we're sharing a lot of info, but where are we going with this?

Dan: we want to get some kind of samples together of different things we're doing in chem
... for example, inline formulas vs set-off equation for the rusting of a metal
... everything you would encounter in a chem book
... that's something I can do

George: can we get the publishers to give to us some samples
... I don't know how many we need
... we could either start with publisher samples, or ask volker to give us some samples that are already worked up
... do we want to put in EPUB? Gotta get it in web page first
... let's collect ten pages iwth chemistry, and start building from there

Volker: dan meant more than molecules... inline formulas, etc

Dan: until we got those, we'll ask "what about X?

CharlesL: should we set up a github at Daisy or Diagram?

George: I think DIAGRAM center would be a reasonable place

CharlesL: I can do that

George: this clearly is an area that needs attention and funding and resources

Volker: since you mentioned funding, since people like what they're seeing, I need to know how can I fund this

Ashley: is there a reason you don't just use international chemical identifiers

Volker: we can use all of the above, we can use ???, ????, ???, ???, and ???

Ashley: if you just use ICI, that would seem to render the answer to how you need to do formatting without the computer having to guess
... once you have a mole file you also have the layout

CharlesL: reading systems would have to figure it out

Volker: we go to SVG

RRSAgent: draft minutes

RRSAgent: make logs public

RRSAgent: draft minutes

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.154 (CVS log)
$Date: 2019/03/26 15:02:22 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.154  of Date: 2018/09/25 16:35:56  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)

Succeeded: s/???/Steve Noble/
Present: janina dauwhe George Avneesh Barrett franco Bill_Kasdorf CharlesL
Regrets: liam
Found ScribeNick: dauwhe
Inferring Scribes: dauwhe

WARNING: No "Topic:" lines found.

Found Date: 26 Mar 2019
People with action items: 

WARNING: No "Topic: ..." lines found!  
Resulting HTML may have an empty (invalid) <ol>...</ol>.

Explanation: "Topic: ..." lines are used to indicate the start of 
new discussion topics or agenda items, such as:
<dbooth> Topic: Review of Amy's report

WARNING: IRC log location not specified!  (You can ignore this 
warning if you do not want the generated minutes to contain 
a link to the original IRC log.)

[End of scribe.perl diagnostic output]