Semantic Identification and Translation Research
Welcomes to our semantic translation project page. If you'd like to
go directly to our demos, click below:
Semantics?
Most mathematical documents are in languages that only
describe how math should be displayed. They do
not describe what that mathematics means (its
semantics).
A primary research program underway in the PolyMath group
studies the issues centred around identifying the semantic
content of mathematical documents, and translating the
mathematics in them into a form that can actually be
manipulated on a computer. The various parts of the
PolyNet research program then look into how we can move
this semantic information across the network and between
tools and environments for scientific research and
education.
There are various research projects, at other Centres,
which study the language for transmitting
semantically-encoded mathematics. Primary examples are
OpenMath
and
MathBus.
These languages, and most symbolic math systems, use representations based
on LISP-like directed acyclic graphs (DAGs). Thus our projects concentrate
on the storage and transmission of such graphs using
distributed
object networking and the building of these objects directly from
the non-semantic information contained in mathematical documents
written in languages like LaTeX. Our systems can use any semantic
language, but we have concentrated on implementation of OpenMath-based
prototypes.
Our semantic identification and translation projects are the
following:
- LaTeX to OpenMath Expression translation
Where we take standard LaTeX documents and
identify the mathematics contained in them. The
resulting mathematics is then translated into
Openmath. We have a fully functional prototype,
written in WildLife, a language used
in artificial intelligence research. You
can try it out via our LaTeX
translation applet
- OpenMath to Everything
Where we build up dictionaries that convert
OpenMath into various important forms (e.g.
Maple, plain ASCII text, and LaTeX). This
allows us to feed the OpenMath directly into
sophisticated mathematical tools. We have
two prototypes developed:
- A Wildlife OM2Maple, developed by
Paul Irvine and Petr Lisonek
- A Java-based OM Translator that moves us
towards being able to store OM content dictionaries
as objects in distributed object database and
computing systems, and being able to dynamically
extend them as needed. A prototype has been developed
by Petr Lisonek.
You can see an applet-based front-end demo
of the OpenMath translation system.
We also have a version that
feeds Maple directly.
SPB