Partnering Plans
Over the past few years, I have built up a collection of software for
formally modeling the intended meaning of English words, phrases, clauses, and
paragraphs. With suitable help, I am now packaging it into useful services that
Lexikos Corporation can commercially support.
By exposing to other software the intended meaning of English inputs,
this technology will facilitate the I/O of English documents for indexing
or semantic processing; help transcribe voice into software-meaningful inputs;
and support multimedia interfaces for man-machine dialogs.
Typically, our modules will be custom assembled as linguistics components in
larger computer applications, and for this reason, I expect them to often be
used in markets served by larger technology firms. Our English-savvy I/O
modules will enhance other software to benefit its end-user, and ultimately all
involved. Lexikos' sales efforts therefore focus on finding good partners
whose products and customers would gain from such a union.
Executable Words
One basis for such partnerships is the dependence of all semantic
technology on domain-specific lexicons and ontologies. These are atypical software, yet
they must be at least partly built before related applications can work. They also
take considerable care, for vocabulary models inescapably limit both the range of English
topics that can be grasped by a system, and its depth of reasoning about them.
The need for such complex models usually implies that the contextual domain of
each discourse be bounded. Humans also face these limits in our natural
languages, but (one domain at a time) we each push them back. With superb innate word-learning
skills, plus time with teachers and training materials, people assemble, debug and expand the
computing objects we so casually cite and share as English word meanings.
Lexikos' semi-automated word-learning approaches operate faster, with a tighter focus and clever tricks,
but they are not totally dissimilar. Vocabulary assembly still takes time and effort - plus special skills
to get it done usefully and efficiently. Some word-learning can occur "on the job", or be reused across
domains. Both tricks help cut net R&D costs, and so does Lexikos' experience in this specialized area.
Domain Lexikons
It is wasteful and counterproductive for every organization wishing to
pursue semantic technologies or English understanding to separately assemble
their own lexicons from scratch, or struggle to integrate "free" ontologies built
for disparate goals by ad hoc methods. The Lexikos value proposition lies
partly in centralizing such work to simplify, regularize, and minimize it, and
also in our well-tested baseline of lexical data and utilities.
As this site shows, we already have assembled broad-but-shallow core vocabulary,
plus semantic parsing tools built to exploit it. Separately, we have other
software able to expand this core for client-specific objectives, in which bulk
lexicons of domain concepts can be manufactured with machine assistance,
rather than hand-crafted at a keyboard.
Let Us Help
If this all sounds a sensible approach to helping your applications
or customers acquire domain ontologies, semantic models, and/or formal-language
inputs for reasoners, please contact me. A nice partnering model probably
is lurking nearby, able to meet your needs sooner, better, and at lower costs, and
I'd be eager to help uncover it.
|
