Suggestion about generators of MathML code

MathML was developed to facilitate the transfer and re-use of mathematical
content between applications.

However, since each tool generates|understands completely different code
we cannot develop a generic parser receiving content MathML code from
several authors using different software.

The problem increase with feedback. Take authors J send us fragment of
content MathML next is checked or simbolically evaluated and output
returned to J. Depending of the MathML software J is using, she|he will be
able to open the content or no.

I could develop an one-to-one tool comparison and generate specific XSLT
templates for each MathML fragment. This is highly expensive and time
consuming but the main problem is that do not work with multiauthored
docs, containing content MathML fragments from different tools. Since each
fragment is not identified, templates cannot be applied because rules are
tool specific. for instance in some cases we wait eliminate <mi> in others
we wait introduce extra <mi>, etc.

Even the simple examples introduced in

[http://lists.w3.org/Archives/Public/www-math/2006Jul/0114.html]

1) a + b

2) sin &pi;

3) -5

4) &int; sin &omega; d&omega;

5) 3/4

6) sqrt(x)/(y^2 -1)

7) -x

8) &int;_a^b &omega; d&omega;

9) x >> 0

10) <p>My favourite Greek letter is &beta;</p>

11) x_i = 5

12) {}^7log x

13) (x+3)^2

14) a/b; a=3, b=4

15) 123/456

are parsed with great difficulty. Take the first example, i would wait
something like

<math>
  <apply><plus/>
    <ci>a</ci>
    <ci>b</ci>
  <apply>
</math>

transforming it to Scheme (+ a b) for symbolic evaluation, but oops i got
error because the file i received from HERMES generated the s-expr

(+ <mi>a</mi> <mi>b</mi>)

All of above examples are extracted from current sites on MathML. Next is
of interest to the Center

16) (&partial;&rho; / &partial;t) = L &rho; + &epsilon;(&rho - &rho;_0)

The RG-1 can work in the research of suitable relativistic expressions for
the L superoperator (Sch-KG, Dirac, or R-QFT propagators are plain wrong
since compact support for wavefunctions is not maintained -the current
research tendency is to obtain propagators are cuadratic in momenta-).

RG-2 can work in the research of generalizations of Abe kernels for the
Zubarev term &epsilon;(&rho - &rho;_0) [*].

If RG-1 and RG-2 are using different MathML software the communication can
be difficult or even impossible.

Proposal]

Use an attribute informating of the profile and other informing of the
generator. For instance

<math generator="HERMES" profile="normal">
  ...
</math>

and i would apply special XSLT to the nodes generating the desired (+ a b).

<math generator="ConText" profile="TeX-annotation">
  ...
</math>

and i would apply special XSLT to the nodes introducing 'lacking' <mi> and
<msub>.

[*]
[http://www.scielo.br/scielo.php?pid=S0103-97332005000400017&script=sci_arttext].
No MathML, GIFs.

Juan R.

Center for CANONICAL |SCIENCE)

Received on Thursday, 27 July 2006 11:04:48 UTC