Now this is impressive

[mmcirvin]

Nth in a series of breathless posts about Wikipedia:

I don't know whether Wikipedia will survive long-term as a repository of human knowledge, but anything that automatically parses TeX markup for mathematics can't be all bad. This explains why there are so many equation-covered pages in the physics section; many people in the business have TeX down so cold that they can type and read equation markup in an almost conversational mode. I'm rusty, but there was a time when I was pretty close to that (though switching between TeX and Mathematica could be jarring).

Several years ago, this fact created confusion on the sci.physics newsgroups on Usenet. Physicists and physics students there made such heavy use of informal pidgin-TeX to convey mathematical formulae that many non-physicist readers became convinced that there existed special newsreaders unknown to them that were capable of rendering the equation markup inline. It was almost impossible to convince them that this crazy-looking markup was actually being parsed directly in people's heads.

(To answer my first question: the ALT text for Wikipedia's generated images is the TeX source. I think you could do better than that, but it wouldn't be so easy. The question of accessibility for the blind and visually impaired on math-heavy Web pages is something I've struggled with for a long time. In theory, MathML markup plus easily available intelligent renderers would solve the problem, but that in theory part is quite a qualifier.)

Comments

[jvandenberg.livejournal.com]

The problem is that Maths notation is essentially non-linear. I mean, look at

\int_0^\infty \frac{x^2+4^x}{xj_0 + x^{1/2}j_1}e^{\sqrt{x}} dx

Reading maths out loud is ambiguous. I think it sucks. However, all the linearisations of notation are annoying to read, like TeX, or make assumptions about the structure of the notation, like MathML (also too verbose).

I however don't have an alternative. It would be nice if I did. E.W. Dijkstra had a farily neat system, of using angle brackets to represent sequences, and then putting the folded operator just after the open brackets. And his function notation I liked.

Actually, I wonder if this is more readable:

int_(x,0->inf), ((x^2+4^x)/(x*join_0 + x^(1/2)*join_1 ) * exp,sqrt,x)

where f,x means function application, and is right associative. We also get unlimited length names because concatenation no longer means multiplication.

Or maybe we could go lispy.

(integrate 'x 0 'inf (* (/ (+ (^ x 2) (^ 4 x) blah this sucks I give up.


Anyway, I should be doing a differential equations assignment.

[mmcirvin.livejournal.com]

MathML is a typical W3C project: crippled by excessive ambition and unwillingness to leave anything out. They couldn't decide whether it was just a mathematical typesetting language like TeX, or a descriptive notation capable of fully characterizing a formula to a machine, like Mathematica's or Maple's language. So they added in both, so that the full language is implemented by nobody, and then they applied XML in such an elaborate manner that neither is really human-typeable in the way that TeX and Mathematica are.