Brainhat is an NLP executive that
can dispatch tasks, handle natural language events, evaluate
ideas, and ask and answer questions. Because Brainhat is knowledge
representation-based, it can digress within a conversation, answer
questions outside the script, and keep rich context.
Brainhat has built-in network
capabilities, making it a clearing house and consolidation point
for natural language events and delegated tasks.
Applications for Brainhat are
limited scope, prepared sessions, in many ways similar to the
statue or restaurant demos you will find elsewhere on this web
site. Personal assistants, kiosks, switchboards, and help desk
applications are other examples. And Brainhat can be extended to
act as a natural computer interface for a robot or other program.
A "brain hat" would, to that end, be a wearable computer
that talks to you.
People interact by talking, and
they mull things over using natural languages, such as English.
But it's different when people interact with computers: commands
are given through some interface and are broken down into
execution primitives. This is true even in most natural language
systems; though the user's intention may be harvested using
language, the result is low-level computing.
Why should it be this way? For
many of the kinds of processing we envision for natural language,
the appropriate level of computation is best modelled by the
self-talk we engage in when we are thinking through a problem.
When searching for an answer, the computer may ask itself
questions or ask questions of others, including other instances of
the program--all in English. Only when it is time to leave then
natural language space, such as when we are directing a robot to
move, does it make sense to drop computation to a lower level.
Brainhat is available in binary
form. Somewhat outdated source is available too. There will be no
charge for academic, personal or commercial use for all version
1.x code. Please see licensing information
for more details.