ORB Visualization
(soon)
National Knowledge
Project
Open question in knowledge representation
Paul:
I think this notion of aligning semantic (entity-relationship) networks with natural language syntax patterns is now centuries old and the heart of Peirce's approach to semantics. John Sowa has added a great deal more to this and is still most enthusiastic about the success of his own approach to natural language understanding.
I do not think syntactic alignments have any success with the semantics problem, nor with non-linguistic theory representation. This non-linguistic theory representation is the deeper n-dimensional (informational) science problem that I think is my most preeminent concern.
Language successfully communicates agreements between individuals both of whom have already a shared understanding of the concept description -- semantic meaning match ups. Language does not create these match ups, which are completely dependent upon shared levels of education and successful mutual assumptions of purpose and intent, things nowhere guaranteed by any linguistic source document taken alone.
If intelligent, but unprepared human scholars cannot divine clear meaning from language representations, why assume that machines will manage the task with far less experience and creative inspiration?
My prior suggestion still holds and will be further detailed in the coming Seybold article by Mills Davis. Getting machines to read our ambiguous language representations is the constant red herring. Language to semantics is a "reverse process".
The semantic web is the unambiguous representation we must jump to first, from there international natural language ambiguities can be (better) resolved subsequently with "forward semantics to linguistic processes."
Dick Ballard
Knowledge Foundations Inc