|
Synopsis
Deepti -- A Hindi Speaking Chat
Robot.
" If anyone ever wanted to merge expert systems, user
direction, and process control here's a potentially hot AIML ticket
to do so! "
Check
the News
for all the updates
Project
Deepti in NEWs: Check out the following links
BBC
World Service: http://news.bbc.co.uk/1/hi/technology/2209775.stm
Rediff
News: www.rediff.com/search/2002/sep/02bot.htm
Siliconeer:
http://www.siliconeer.com/past_issues/2002/september2002.html
History
Communication must have been one of the most complex
art or science, whatever we might call it, practiced and developed
by humans. Exchange of ideas and thoughts make (and always did)
possible the accumulation of knowledge, which is certainly the
single most important factor for the advancements that we are always
proud of.
One thing that is done for most of human seconds
around the world is, of course, chatting (of course in different
flavors, "talking", "discussing", "debating", "delivering", as you
like it). Though purists would not shore this view but we hold that
one can't chat to "himself". Thus we all need to be chatting to
"someone". But humans come with a host of attributes, have their
constraints, limitations, are moody and tend to get irritated when
told something they don't relish. This was an observation recognized
long ago. We have had movies exhibiting a sensible talk between men
and machine. That is really exciting because fortunately the machine
is built devoid of the limitations that exist by virtue of human
nature. A machine would talk to us as and when we want it to.
Humankind really graduated on this path, when some
really talented computer scientists set out to find an
implementation that is no longer mere fiction. Many such efforts
have been made and in the present context we are most concerned
about the most simple of Natural Language Processing techniques and
also undoubtedly the most successful in recent times. Its based on
pioneering work of The AI Foundation in developing an as yet
primitive non-standard "markup" language for Artificial
Intelligence, called the obvious, AIML.
There have been rave claims by the Special Interest
Groups (and the pool of extremely
talented people who are working very hard in making AIML an
accepted standard) of the application of AIML. Attempts have been
made to develop chat robots using this, and a respectable degree of
success has also been achieved.
AIML is the brainchild of Dr. Richard S. Wallace,
PhD. Computer Science, Carnegie Mellon
University.
What
we want to do?
The project would be an attempt to realize the fancies
that spring up after reading our project title. We hope to come up
with a chat robot that understands and speaks Hindi and is capable
of intelligent talk.
Why we want to do this?
It took only a page and a half of our AI text book to
entice us into Natural Language Processing with the classic example
"I want to print Bill's .init file". Thus this project.
We sympathize strongly with the AI Foundation and applications
using AIML take the top slot in their list of ongoing research. The
work and thoughts of Dr. Wallace are undoubtedly too tempting to
resist. Thus this project.
It would be redundant to say that AIML would be a
premier NLP technique of the times to come. Thus this project.
Also we have all through been sympathizers of Open Source Foundation, thus pursuing
their cause has always been our endeavor. Thus this project.
How we intend to do it?
The nature of the project renders itself best to
modular programming. We wish to keep "From Simple to Complex" as our
motto throughout the development process. We would wish to set out
with a study of AIML along with planning and discussing the simplest
of the AIML "categories".
The development and conception of these categories
will undoubtedly be a continuous process.
We would need to develop what might be called an AIML
Server, that understands the AIML rules and hence our AIML
categories. This is actually what the "robot" is. With a weak
"shell" developed, we move on the give a simple command line
interface (aka client), which would infact be an interface to the
human user. This is clear if we take our server to be analogous to a
http Server and the User Interface as being analogous to an http
Client that sends requests to the server on behalf of the user.
Until now both the user and the bot communicate
through Unicode (or ASCII) characters as the case would be. At this
stage we would work towards transforming the bot output to simulate
human voice. This can be done using a Text-to-Speech system (TTS).
The development of such a system might also be one of our concerns,
if nothing more fortunate happens. Now we have the user inputs in
text format while the bot "speaking" out her dialogues.
We would like to reiterate that all this while the
categories of AIML would need to be refined to render Deepti
intelligent.
We would also wish to move from command line interface
to a web browser compatible interface. For this we would need to
direct all input to a software port at which our server would be
listening, so that appropriate responses can be played on a
multimedia enabled system through the speakers.
Just in case we have the best of fortunate times, we
would also wish to have some form of lip syncing, so that we have a
character on screen that seems to be speaking out her
responses.
Who we are?
We are a three-member team working along with Rishi. The team members have
varying skill sets and have had varying experiences during the
vocational trainings and projects undertaken over the previous
years. We hope to work cohesively towards our common endeavor.
The team comprises by the following:
1. Ritvik Sahajpal
2. RohitKumar
3. Rahul Jindal
4. Rishi Bihari
Any queries about the project may be directed to any
of the members above or to Zincers
mailing
list |