HPSG Diagrams

[Luis Casillas]

On Tue, Dec 03, 2002 at 02:36:33PM -0500, Martin Jansche wrote:

> On Tue, 3 Dec 2002, kaplan wrote:
>
> > is there any way of typing in a tree-description (e.g. something
> > like a labeled bracketing) and having it pop out to a nicely drawn
> > tree?
>
> There is one for LaTeX -- the standard, venerable, stable, virtually
> bug-free, ubiquitous, easy to use, free, low-cost, multi-platform,
> backwards compatible, no-nonsense, full-featured typesetting solution.

A fair appreciation of LaTeX should not only reveal these true virtues
(of which, however, I dispute the "easy to use" and the "free", the
latter in the basis that the time you spend learning it is not free),
but a fundamental problem: bad separation of the two problem domains
it deals with.  LaTeX is a system that conflates two functions, writing
documents and extending LaTeX, into a single language, and one not
very good for either task IMHO.  As a language for the actual writing
of your documents, LaTeX, being Turing complete, is absolute overkill;
it's also hard to parse.  As a Turing complete language for coding
extensions to your document language, it turns out to be a terribly
underfeatured and hard to use language.  It tries to kill two birds
with one rock, and only manages to scratch a wing from each.

Also, having been designed about 20 years ago, TeX simply does not
interact very well with modern stuff it would be nice to plug it up
with.  The best thing to have would be some sort of system that allows
you to write documents in SGML/XML, to be interpreted by some sort of
compiler that allows you to write plugins to extend the syntax in your
programming language of choice to do the things that LaTeX packages do
(and build up a community that writes such packages, like LaTeX has
around it).  This would allow one to have packages that directly
formatted all sorts of XML data.  Of particular interest for us,
corpora nowadays tend to come in such formats.  You could have modules
that allowed you cut and paste raw XML corpus examples and format them
automatically.  Some of these ideas I believe are being worked on for
LaTeX3, which from what I gather is being designed for automated
translation of SGML/XML into LaTeX.

This is getting long and complex, but my point is that there is plenty
not to like about LaTeX, either if you're a computer-scared person who
thinks WYSIWYG and Minimalism are both intuitive because they both
involve moving items around visually, or if you're a mighty hacker who
thinks HPSG and compilation-based typesetting are the best because they
both involve working with well defined and powerfully expressive
formal systems.

I have a fear that this discussion might be perceived to be straying
beyond what's relevant to this list, so I'll finish off with one
important fact to connect it back: the HPSG conference proceeding
submissions this year (and I presume in earlier years, too) were
required to be in LaTeX.  If you don't know LaTeX and can't find
somebody with the time to help you out with it (or can, but can't/won't
pay them to), then this is a hurdle to jump to publish your work on HPSG.
The same applies I believe to many multi-author volumes produced by the
HPSG community (and the Stanford/CSLI community in general; I coedited a
book where at least one of my authors had to pay somebody to translate
their submission into LaTeX).

--
Luis Casillas
Department of Linguistics
Stanford University
http://www.stanford.edu/~casillas/