Accessible open textbooks in math-heavy disciplines | Hacker News Reader

Accessible open textbooks in math-heavy disciplines

richardzach.org

235 pointsvolemo75 comments

> Switch the language on foreign terms and names so that screen readers can pronounce them in the right voice.

Screen reader user here. Don't actually do this, this is bad advice.

Just like a lecturer won't suddenly switch to a German accent when saying words like "schadenfreude" or names like "Friedrich Nietzsche", neither should a screen reader. Having your voice constantly change under you for no apparent reason is distracting more than anything else.

What you should do this for are longer pieces of text in a foreign language, like a multi-paragraph piece of text to analyze in a foreign language textbook.

Yea, I hate that. Words are pronounced differently in foreign languages. Do we say Moscow or Moskwa? Do we say ka-tana or ka-ta-na? If Freud is not spoken with the typical Gemran diphthong, then suddenly someone comes along and corrects you. I do speak German, I know how Freud is pronounced and I will pronounce it as it should be pronounced when speaking German, but when speaking English, it is Frood for me.

So, I am with you. We shouldn't learn the pronunciation of 200 different languages. If Kirchhoff's laws sound like Captain Kirk, who the fuck cares. Different languages pronounce stuff differently.

i agree with you in spirit (I pronounce Paris as Paris).

however I have never heard of someone pronouncing Freud as Frood, outside of "Bill & Ted's Excellent Adventure".

https://www.youtube.com/watch?v=Sobc2WhL16c

Fair enough. Video doesn't play, but I believe you. I don't know where I heard froodian slip and frood, but I checked a few places where they pronounce it and all agree with you. Bet I will find more example as soon as enough time has passed so it would be weird to post it here. Damn you, Murphy.

I do agree with you in that I say "shampain" instead of "champann" (champagne) when I'm speaking English instead of French. Language is a tool for communication first and foremost.

[Hi from Argentina!] For 'Euler", I keep switching randomly betwen

Eh-oo-leh-r that is how it should be read if it were an Spanish word.

Oh-ee-leh-r that is the proper German pronunciation

And here I felt he added lubrication to machinery = Oiler and my friend Eugene who's mother called him Oygen does the same. Being from the UK, came to Canada in 1948, I spoke colloquial English in school, but correct London Cockney slang at home to family - on phone calls to friends, if I responded to family mid dialog, my friends would always ask who was that when my slang was over heard.

>I pronounce Paris as Paris

there is an S in Paris because the French used to pronounce it that way and it got written down that way in French... and that is also when that word got added to the English lexicon. Paris is a word in English that is pronounced as it is spelled. There is a French word spelled the same way that is pronounced differently. Something similar is true with Moskva/Moscow (btw, people in Moscow, Idaho pronounce it "mosko")

these type of historical borrowings don't offer useful guidance to how Freud should be pronounced in English.

I agree with this way of thinking about it, but the problem is “added to the lexicon” is ill-defined.

There is no official lexicon. When speaking English, the pronunciation of “Paris” has become well-established, but for countless other words, it has not.

linguists use the word lexicon (as opposed to dictionary) to mean those words which are spoken as prevalently, let's say, "as the syntax in which they are agreed and declined". (I just came up with that and think it's quite clever)

it has become over common to over point out that linguistics is descriptive, as if anything goes; anything does not go, and that is what linguists study. Stray from the lexicon, and people will ask what you are talking about. When they stop asking, it's in the lexicon.

> I do speak German, I know how Freud is pronounced and I will pronounce it as it should be pronounced when speaking German, but when speaking English, it is Frood for me.

That... isn't the normal English pronunciation. The English pronunciation would rhyme with "joyed", if "joy" were a verb.

/'sɪg.mənd fɹɔɪd/

There are some other big names where the same vowel sequence isn't recognized: Euler (usually pronounced with /ɔɪ/) and von Neumann (not so much).

Euler suffers from beginning with the "eu", which makes it look more Greek.

> There are some other big names where the same vowel sequence isn't recognized: Euler (usually pronounced with /ɔɪ/) and von Neumann (not so much).

I always have trouble finding a reference for the sounds corresponding to IPA symbols, so I'm not sure what you're claiming for the pronunciation of either of those. But, at least among the mostly American mathematicians I know, the 'eu' in 'Euler' and 'von Neumann' are usually pronounced the same way we pronounce the same way we pronounce the 'eu' in 'Freud' (which I agree is essentially how I'd pronounce the 'oy' in 'joyed').

https://www.collinsdictionary.com/dictionary/english/von-neu...

OK, thanks; that does make the claimed pronunciation clear. Still, I have literally never heard a mathematician pronounce the name that way (or Euler's in the analogous way).

> Do we say Moscow or Moskwa?

Russian friends taught me that there is no "o" (as the letter is pronounced in Spanish or German) to pronounce in Москва since the о is unstressed. Rather pronounce it as "Maskwa" ("a" letter as in Spanish or German). :-)

True to a first approximation. A good second one is that a carefully enunciated [ɑ] isn’t correct, either; a schwa [ə] will sound better, so the vowels and the overall rhythm will be similar to the English word bazaar in a non-rhotic accent. Finally, the hard reality[1] is that all of this is heavily accent-dependent: in Vologda (500km from Moscow) you will hear [o] for the first vowel; in Ryazan (200km from Moscow) you can hear [a] ~ [æ]; even in Moscow itself, a radio announcer will say a fairly careful [ɐ] while someone who grew up in the poorer suburbs will have an almost-inaudible [ə] (this is a strong class marker).

The English word Moscow, meanwhile, is itself very interesting: it’s not actually a derivative of the Russian Москва, but rather a cognate, as both of them are derived[2] from different cases (accusative vs. locative or genitive) of the original Old East Slavic (aka Old Russian, aka Old Ukrainian, etc.) name.

[1] https://en.wikipedia.org/wiki/Akanye

[2] https://en.wikipedia.org/wiki/Moscow#Etymology

> Just like a lecturer won't suddenly switch to a German accent when saying words like "schadenfreude" or names like "Friedrich Nietzsche"

Is there a middle ground? Whenever I check my content with a screen reader, uncommon foreign names are often mispronounced in ways that are sometimes almost irrecognisable. Even my name comes out wrong, although it would be understandable (typically, the stress ends up on the wrong syllable).

A lecturer who is fluently multilingual might indeed smoothly switch accents when pronouncing foreign words. But it's still the same voice, and (if they're well practiced at it) they don't have to pause in mid-sentence to switch languages, as text-to-speech systems usually do. And eSpeak can switch languages while still being the same voice, since it's a rule-based, parametric synthesizer. But, at least with NVDA, a mid-sentence HTML span with a different lang attribute still causes a (short) break in the intonation on either side. That's too bad, because a multilingual parametric synthesizer like eSpeak could be like the ultimate polyglot speaker, impressing us all with how smoothly it switches languages.

Out of curiosity, are there models for speech synthesis which generate speech parametrically and then refine neurally?

FYI, you seem shadowbanned[1] for some reason: all of your comments after the first one (and, now that I’ve vouched for it, this one) are marked “dead” as though downvoted to oblivion, though I find nothing that objectionable in any of them. I suggest you look at your comments page[2] from an incognito window (and perhaps contact 'dang for clarification?).

[1] https://github.com/minimaxir/hacker-news-undocumented#shadow...

[2] https://news.ycombinator.com/threads?id=mpascale00

Fixed now. Thanks for watching out for a fellow user!

What would it mean to say a german word with a german accent? I thought a german accent is when you pronounce english words as if they're german?

Richard Zach et al. logic textbooks are a work of art: https://slc.openlogicproject.org.

Aside, I'd be happy to pay for great classic math textbooks in a well typeset edition.

Finite Dimensional Vector Spaces is one such example. The text is fantastic, but everything is too cramped and hard to read compared to a modern book from the 1970s onward.

A newer edition typeset in LaTeX would be great.

By my estimation, it’ll hit the public domain in 2053. Not that long of a wait, and in any case math typesetting should be much improved by then.

2053 is pretty far. I wish Dover, who hold publication rights, improved typesetting of this and a few other classic textbooks.

It's probably not hard to use an LLM to do the bulk of the conversion to TeX work cheaply, and then some human input to polish the final document and fix errors.

2053 is too far away indeed.

“It’s probably not hard…” - how many such wishful thinking statements were uttered by humanity.

Have you tried image to LaTeX translation? It works fairly well already for text plus equations.

I've done this for a 30-page manuscript with no sources, and I was able to recreate the entire document with minimal manual intervention to get a correct PDF.

This is not programming, it's OCR and translation to a very simple markup language. It's a very easy mechanical task.

There was a project a while ago to do exactly this (in general, not for Halmos's book specifically. I assume that's the one you mean). I can't find it by Googling, but here's an Internet Archive capture of a Project Gutenberg page that collected books for which this had been done: https://web.archive.org/web/20120502091427/https://www.guten... .

Typst( https://typst.app/) Is good Latex alternative which should be mentioned here. Their roadmap also features html as a target.

HTML is available as a preview feature now

Switched from LaTeX to Typst two years ago and have never looked back. Looking forward to their HTML export feature.

Looks like you can already play with it (though it's still "very incomplete") https://typst.app/docs/reference/html/

A friendly person on the internet already put up the typst-to-mathml part of this [0]. I have been considering the ultimate yak-shave of building a static site generator around this...

[0]: https://github.com/wcshds/typst-math-to-mathml-converter

Another is TeXmacs [1].

"Despite its name, TeXmacs is not a front-end to TeX or LaTeX.[mHowever, TeXmacs documents can be converted to either TeX or LaTeX. LaTeX also can be imported (to some extent), and both import from and export to HTML, Scheme, Verbatim, and XML is provided; the HTML export is stylable with CSS (since version 1.99.14). There is a converter for MathML as well, and TeXmacs can output PDF and PostScript for printing."

[2] GNU TeXmacs:

https://en.wikipedia.org/wiki/GNU_TeXmacs

Yeah they really need to rebrand.

If only it weren’t subscription model. At least I can still install LaTeX without a credit card.

Typst is Apache licensed. I install it with "cargo install --locked typst-cli". The subscription model is for their web interface with storage. I write simple worksheets for a living, so I've been using it instead of TeX lately. (I've use LaTeX or plain TeX since around 1990.)

That "LaTeX alternative" looks awful? `integral_t^oo` in place of `\integral_t^\infty`?

Why would anyone switch from LaTeX to this other than the speed?

Its a lot easier and idiomatic to write. I would try it first before knocking it off.

> Its a lot easier and idiomatic to write.

Surely both of these characterizations depend on the person? I can believe that integral_t^oo is idiomatic if that's what you're used to, and maybe it's easier to pick up from scratch, but, for someone long used to TeX, it just makes me wonder what other unpleasant surprises someone else will have decided are actually pleasant.

My current favorite option is Quarto [1]. It's basically a friendly wrapper around Pandoc [2], letting you write in Markdown (+ lots of Quarto-specific extensions) and render to LaTeX, Typst, multi-page HTML, EPUB, docx, and more.

[1] https://quarto.org/

[2] https://pandoc.org/

It's mentioned in the article.

I think the neatest feature would be: the mathematician states the theorem in English (edit: with math symbols as needed), then he states the proof in English.

A post-Processor transform the proof into lean (with any official lib loaded). It is automatically verified. If something is missing, the post processor ask to write in English the missing parts. Iterate like this.

The lean proof is hidden in the final document, and can be displayed if needed. Or even, we get an English version that can be easily retransformed into lean at will.

Bonus point: we can query the document to give more details on part of the proof and it outputs (expanded) lean formatted as nice English.

Note: there is no need to have all the math self contained in the document, he can say to assume some theorem true to do his proof. And this would be reflected in English.

I can’t speak for true mathematicians, but as a (mostly theoretical) physicist I would hate to write my maths in plain English. Notation is used precisely because it’s easier, more efficient (to write and to read) to use it instead of prose. If you’d like an example, I suggest reading texts on solving quadratic equations that predate modern algebraic notation.

Sorry I meant English with maths inside, not to avoid all symbols! (A classical style proof that you find in any current paper) Edited.

Another LaTeX to MathML option that I have had a good experience with is temml — it works in the browser or server side with Node.js. https://temml.org/

For the articles on my website, I have a pretty janky workflow where I write a LaTeX document that I compile both to a PDF and (using Pandoc) to HTML, which I render with KaTeX. I've been in the market for a while for something that's less fragile but which can still produce both a PDF and visually appealing HTML output starting from a LaTeX source, and it seems like some of the ideas listed here might be what I want! Thanks for the link. (That said, if anyone has a particular recommendation, I'd love to hear it!)

I’m lazy. I just extract the abstract from the tex source and put a link to the PDF. That is however quite different and not public in my case.

Turns out from a technical documentation consumer perspective it’s easier to save a PDF than feck around with web pages and try and save them.

Accessibility is win-win because it helps more people access information and also has content creators thinking clearly about the quality of code and presentation.

Why do maths textbooks never explain how to read the equations? I've tried reading a few and before you know it there'll be some matrix stuff in, integrations or whatever and no explanation of how to read them in normal English like a maths teacher does.

They all seem to assume you're still at school rather than it being decades since you last had to read them. Even using Greek symbols - WRITE THEM OUT. Jeez, it's so elitist.

Try reading physics books that introduce the maths as an instrument (e.g. Wald’s General relativity to learn tensor calculus, you only need the first few chapters), they usually explain concepts and notation in simpler terms. However, they may hide stuff under the rug of “it’s a tradition”, “we do this, don’t ask questions”, “ margin is too narrow to contain the proof”, etc.

P.S.: Greek alphabet is not that bad (it’s only 24 letters, and a third of them isn’t used on behalf of being the same as some Latin ones), just look it up. It’s much worse when some mathematician in the 19th century invents his own squiggly to refer to a pretty useful function.

If you don't know how to read them, you aren't ready to understand them. The writers expect you to have a base level of knowledge.

At my age I find this issue annoyingly predictable -- it suggests replacing LaTeX with something more modern and/or flexible. But that was the original idea behind LaTeX -- to replace what preceded it, create a portable comunication medium for mathematical expressions.

Not to discourage experimentation, but I would like to see some behavoral reserve and healthy skepticism before adding another layer to the mathematical expression enterprise. There's also this issue to consider: https://xkcd.com/927/ .

I agree, and the article talks about working with LaTeX and converting directly to HTML. This is more about replacing the final PDF with HTML.

But you're discouraging experimentation with nothing else but a reference to a silly comic

The annoying comic would be silly if it weren't true. The point of the comic is the fact it alludes to, which by magnitude transcends any specific reference.

Except it's not true, it ignores the most important factor - market share. If the new 15th standards becomes the most/only used one (as has happened with many new standards), then the silly idea from the comic "the future is no better with just +1 in the pile" doesn't reflect reality, reality would be that indeed that new universal standard solved the problem and reduced the ridiculous fragmentation.

That's why there is nothing healthy behind this thought-terminating comic cliche, just a generic conservative discouragement of any experiments.

What LaTeX replaced was:

a) typing the text of math papers and writing in the formulae by hand.

b) word processing which was fixed width and line based, where there would be three characters for 'top of an integral sign', middle and bottom, and you would have to align your three characters at the same horizontal point on three successive lines to make the integral sign, and also position all subscripts, exponents yourself.

The fact that LaTeX had an important mission and was successful in it doesn't mean that another thing is not needed now. Things have moved on. Replacing things which solved a problem doesn't mean going back to the situation before they existed.

Uh ... where is it written that LaTeX can only provide pdf output?

Turn that around. Thought experiment. Say we work like pigdogs for 5 years and we completely replace all of Latex except the part that produces pdf. What value have we provided to users at that point?

The article talks about converting LaTeX to HTML, which is feasible today, if only buggy and fragile. This is the textbook the author talks about, which is written in LaTeX (but compiled with LaTeXML instead of pdflatex): https://forallx.openlogicproject.org/html/

It does. But why can't LaTex produce png or something? Why does it have to be either pdf or pretty much abandon the idea of typesetting? Or am I misunderstanding?

You can. Compile to DVI then use dvipng. There's not much reason to do it anymore. PDFs take up less space, you can copy the text, do hyperlinks.

So what's so bad about pdfs?

Some people like text which they can reformat easily (increase font size, change to different screens dimensions).

What is generally bad about the PDFs that Latex produces (and is a problem with latex, not a problem with PDF) is that they are very inaccessible, they don't work with screen readers.

The reason it's so hard to make latex output HTML (although people are working on it) is that latex is actually a programming language, which is executed to decide where things go on in the PDF.

Make latex output HTML is a bit like trying to take (say) a game engine like Unity, and change it's rendering engine to output HTML instead of graphics -- in the worst case it's basically impossible, as the game just generates commands like "draw triangle here", without context or semantics.

Even after properly tagging your PDF and using Acrobat Reader to reflow the text, you cannot achieve this level of flexibility: https://forallx.openlogicproject.org/html/

Note how Richard's book adapts to any screen size, can change fonts and color schemes, system settings such as 'high contrast' will affect the rendering of the page, and you could even use browser extensions to restyle the page to e.g. use a more dyslexic friendly font of your choice.

This kind of functionality is not afforded by Adobe Reader. Even the official Adobe's example of reflowing that was posted in another thread is quite bad: https://helpx.adobe.com/uk/acrobat/using/reading-pdfs-reflow... The reflowed PDF is just stacking all text and removing all non-text visual cues. For example, pairs of name/role are separated by whitespace in the PDF, but after reflowing they are undisguishable from each other (who would be the senior VP, Sunny or Daniel?). In HTML, reflowing would preserve semantically relevant whitespace out of the box.

> Compile to DVI then use dvipng

or dvisvgm, which will produce scalable images. However, images are even worse than PDF in terms of accessibility, which is what the article is talking about.

By the way, LaTeX and TeX existed long before PDFs were ubiquitous. A common workflow in the 1990s and prior was

TeX/LaTeX -> DVI -> PostScript -> printer

And DVI stands for "device independent", so the idea was you can take a DVI and convert it to any format. PDFs just eventually became the dominate format.

Yeah, I remember.

I feel like there are enough free and semi-free textbooks to study just about anything I want. The limiting factor is my not so plentiful brainpower to actually learn new math stuff. There's no way I can possibly run out of textbooks.

"BookML: automated LaTeX to bookdown-style HTML and SCORM, powered by LaTeXML" https://vlmantova.github.io/bookml/

LaTeXML: https://en.wikipedia.org/wiki/LaTeXML :

LaTeXML emits XML from a parsing of LaTex with Perl.

SCORM is a standard for educational content in ZIP packages which is supported by Moodle, ILIAS, Sakai, Canvas, and a number of other LMS Learning Management Systems.

SCORM: https://en.wikipedia.org/wiki/Sharable_Content_Object_Refere...

xAPI (aka Experience API, aka TinCan API) is a successor spec to SCORM for event messages to LRS Learning Record Stores. Like SCORM, xAPI was granted by ADL.

re: xAPI, schema.org/Action, and JSON-LD: https://github.com/RusticiSoftware/TinCanSchema/issues/7

schema.org/Action describes potential actions: https://schema.org/docs/actions.html

For example, from the Schema.org "Potential Actions" doc: https://schema.org/docs/actions.html :

   {
    "@context": "https://schema.org",
    "@type": "Movie",
    "name": "Footloose",
    "potentialAction": {
      "@type": "WatchAction"
    }
  }

That could be a syllabus.

ActionTypes include: BuyAction, AssessAction > ReviewAction,

Schema.org > "Full schema hierarchy" > [Open hierarchy] > Action and rdfs:subClassOf subclasses thereof: https://schema.org/docs/full.html

What Linked Data should [math textbook] publishing software include when generating HTML for the web?

https://schema.org/CreativeWork > Book, Audiobook, Article > ScholarlyArticle, Guide, HowTo, Blog, MathSolver

The schema.org Thing > CreativeWork LearningResource RDFS class has the :assesses, :competencyRequired, :educationalLevel, :educationalAlignment, and :teaches RDFS properties; https://schema.org/LearningResource

You can add bibliographic metadata and curricular Linked Data to [OER LearningResource] HTML with schema.org classes and properties as JSON-LD, RDFa, or Microdata.

The schema.org/about property has a domain which includes CreativeWork and a range which includes Thing, so a :CreativeWork is :about a :Thing which could be a subclass of :CreativeWork.

I work with MathJax and LaTeX in notebooks a bit, and have generated LaTeX and then PDF with Sphinx and texlive like the ReadTheDocs docker container which already has the multiple necessary GB of LaTeX installed to render a README.rst as PDF without pandoc:

The Jupyter Book docs now describe how that works.

Jupyter Book docs > Customize LaTeX via Sphinx: https://jupyterbook.org/en/stable/advanced/pdf.html#customiz...

How to build the docs with the readthedocs docker image onesself: https://github.com/jupyter-book/jupyter-book/issues/991

ReadTheDocs > Dev > Design > Build Images > Time required to install languages at build time [with different package managers with varying performance] https://docs.readthedocs.com/dev/latest/design/build-images....

The jupyter-docker-stacks, binderhub, and condaforge/miniforge3 images build with micromamba now IIRC.

condaforge/miniforge3: https://hub.docker.com/r/condaforge/miniforge3

Recently, I've gotten into .devcontainers/devcontainers.json; which allows use of one's own Dockerfile or a preexisting docker image and installs LSP and vscode on top, and then runs the onCreateCommand, postStartCommand

A number of tools support devcontainer.json: https://containers.dev/supporting

Devcontainers could be useful for open textbooks in math-heavy disciplines; so that others can work within, rebuild, and upgrade the same container env used to build the textbook.

Re: MathJax, LaTeX, and notebooks:

To left-align a LaTeX expression in a (Jupyter,Colab,VScode,) notebook wrap the expression with single dollar signs. To center-align a LaTeX expression in a notebook, wrap it with double dollar signs:

  $ \alpha_{\beta_1} $
  $$ \alpha_{\beta_2} $$

Textbooks, though? Interactive is what they want.

How can we make textbooks interactive?

It used to be that textbooks were to be copied down from; copy by hand from the textbook.

To engage and entertain this generation.

ManimCE, scriptable 3d simulators with test assertions, Thebelab,

Jupyter Book docs > "Launch into interactive computing interfaces" > BinderHub ( https://mybinder.org ), JupyterHub, Colab, Deepnote: https://jupyterbook.org/en/stable/interactive/launchbuttons....

JupyterLite-xeus builds a jupyterlite static site from an environment.yml; such that e.g. the xeus-python kernel and other packages are compiled to WebAssembly (WASM) so that you can run Jupyter notebooks in a browser without a server:

repo2jupyterlite works like repo2docker, which powers BinderHub, which generates a container with a current version of Jupyter installed after building the container according to one or more software dependency requirement specification files in /.binder or the root of the repo.

repo2jupyter: https://github.com/jupyterlite/repo2jupyterlite

jupyterlite-xeus: https://jupyterlite-xeus.readthedocs.io/en/latest/

> The obligatory TeXmacs mention.

It is only kind of true, PDF does everything that HTML does, with print quality, naturally one needs Adobe proper for the full feature set, as most readers only implement the standard partially, for various reasons.

The accessibility and reflowability of HTML content, not to mention the ability to customize color schemes, fonts, line spacing, and similar are not possible with PDF, even using Adobe software. Even using the latest PDF 2.0 standard, you are ultimately expected to convert it to HTML if you need all that flexibility (such as via https://ngpdf.com/).

I feel we are splitting airs here, and there is the whole FOSS versus Adobe thingie,

https://helpx.adobe.com/acrobat/using/reading-pdfs-reflow-ac...

I posted this in another thread

> Even the official Adobe's example of reflowing you posted before is quite bad:

> https://helpx.adobe.com/uk/acrobat/using/reading-pdfs-reflow...

> The reflowed PDF is just stacking all text and removing all non-text visual cues. For example, pairs of name/role are separated by whitespace in the PDF, but after reflowing they are undisguishable from each other (who would be the senior VP, Sunny or Daniel?). In HTML, reflowing would preserve semantically relevant whitespace out of the box.

Getting PDF up to WCAG standards is a painful and arduous task.

Reply on news.ycombinator.com