[om-list] project differences
MrM0j0r15n Rajeev the kanehbosm
electrickanehbosm at hotmail.com
Sat Aug 31 02:30:09 EDT 2002
Hello Mr Call and other friends,
As you pointed out, documents are a primary datum for my project. But, i
would like to add that documents are only a primary input data for the KB.
At first, the KB only returns documents from within the repository, since it
is not able to provide NL answers to NL questions that are posed. The idea i
was putting forward, involves a scheme where an intelligent agent would
report results in NL format, but allows users to access the base documents
(if any).
I doubt it will be very possible to build a fully NL solution in the near
future. But, i think its possible to begin the process, by designing a
framework, where the robot is just another human being - able to assimilate
arbitary pieces of information from the KB like other users, and able to
chat (literally) with other users of its experiences.
Let us assume that such a robot is a human being. Once the basic ideas of
language are embedded into the robot (in the form of an NL semantic
analyzer), and a vast ontology is available which is the basic idea of
knowledge, then the robot may create ideas of its own, by expressing them in
NL. However, the problem with this are:
1) a NL semantic analyzer/classifier/generator has not been built
2) a vast ontology + instances of ontological items (ie, Bird is ontology,
knowledge about Bird is instance) is not available, so that the robot is
able to learn not only language, but also acquire knowledge while doing so
(this based on the assumption that knowledge increases language, and
language increases knowledge of knowledge)
Part 2 can be built today. Part 1 may be designed in part today.
insanekane
PS: if any of the term have been misused pls tell me. Also, i di not recieve
Mr Call's message via hotmail. Does anyone know how i can fix it ?
_________________________________________________________________
Send and receive Hotmail on your mobile device: http://mobile.msn.com
More information about the om-list
mailing list