NLP and XTM: a Winning Combination
People have been trying to make computers understand
English for a long time.
One popular approach is to try mapping every sentence into FOL.
That is not a bad idea, but it is hard, as FOL is more rigid that
English and less expressive.
MODELER splits this hard process into stages, and simplifies it by
transforming each given English paragraph into isomorphic structures in
a Topic Map. Client software can then "understand" what was said in
terms of the equivalent network of topics, roles, occurrences, and
associations
created by this process:
- A user of WORDS (human or digital) maps a paragraph into scripts
- MODELER turns them into data in its TM engine (known as CONTEXT)
- XTM equivalents to the paragraph emerge from MODELER as output
- A client loads this XTM into its own TM engine to query its meaning
Our core design premises
are that XTM produced like this under MODELER's idealized English
ontology can record and transmit clear, elegant semantics for
any sentence; and expose them to any J2EE application. All it
takes is using our scripted XTM lexicons to define MODELER's
domain-specific vocabulary.
Testing Our Design Premises
Lexikos wants to test our design and speed R&D on MODELER 1.0. So we hereby appeal to its future user community
for help. We need a few paying, XTM-savvy beta customers now for pre-releases
in which:
- Each input sentence is invisibly scanned by using our syntactic lexicon
- New semantic lexicons then re-express its meaning as commented LTM
This
core process is what we most need to test and refine. The output
LTM acts both as a measure of MODELER's performance and a handy way to
make it smarter. You can beta test it with us, get early access to a new
way to build agents, and price breaks on a prepaid, first year site license.
Why You Should Help: It Creates a New TM Use Case
MODELER is a long
term mission for Lexikos with many elements. We succeed
when it
meets your intuitive desires for a good linguistic U/I module, starting with
NLP-LTM interfaces that indeed can cleanly capture English semantics.
If you help us to refine these interfaces, then besides a fast new way to build your TMs using idealized
English paragraphs, you also get two early business benefits that will eventually spread to the whole TM community:
- Seeing how NLP can focus and simplify back-end TM-FOL interfaces
- Helping make TM paradigms the new standard for an NLP-FOL bridge
MODELER
seeks to use Topic Map paradigms as a new intermediate between two
fast-growing market forces - the NLP world (including voice
processing), and that of FOL (including OWL, DAML, etc). If TMs
can indeed be good middlemen, and bridge long-standing gaps between NLP
(grammar) and FOL (logic), that will help the entire Semantic Web. But
it will especially help the major players in Topic Maps, whose skills,
products and services will become more in demand.
That is ultimately why these
NLP-XTM interfaces have general business benefits to you, and why
Lexikos hopes to get active support from all major players in the XTM
community on beta testing WORDS and MODELER.
Dan Corwin, CTO
|
