[Project-ideas] Gsoc-2012 Project

[Abhishek Gupta]

Hi Abhishek,

I have some comments which I have added inline.

> My understanding of the idea is that we have to process query so that we
> can
> > get the desired result. The query may be in different languages and the
> > program needs to understand the query as appropriate as posssible in any
> > language and give the answer accordingly by searching different
> forums,FAQs
> > or articles like wikipedia.


I feel that this approach makes this as a problem of information retrieval
which it is not. This is primarily, because we should be given some data
which should be processed in the pre-processing step. An example of that
can be found by searching "virtual chat paypal".


> > If the query is interpreted in right way then
> > there are high hopes for giving as appropriate answer as possible. Most
> of
> > the problems in query interpretation may be removal of  some unwanted
> words,
> > pointing out the key words, recognize the words in the right sense
> (remove
> > disambiguation) and recognizing the grammar of the query.
> >  I am interested in working for the disambiguation and noise removal in
> the
> > query. So, I would like to know how much has been done in that direction
> and
> > what can be further done???


So, I feel that the large parts of the project might involve working on
other aspects like knowledge representation of the input data, making your
system to ask intelligent questions in case if it is not sure about the
question (read about "interactive searches"). Disambiguation shouldn't be
the primary problem because it generally involve semantic analysis and
given the scope of the system like say a FAQ page of a website with about
3000 lines of input data, there shouldn't be much ambiguity within it.

At the work done, we can probably follow the publications with SIRI which
is a derivative of the work done in CALO ( "Cognitive Assistant that Learns
and Organizes") project. They have quite a large number of publications
which can be found here - https://pal.sri.com/Plone/publications.

Best Regards
Abhishek
abhishek.cc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ankur.org.in/pipermail/project-ideas-ankur.org.in/attachments/20120323/732df236/attachment-0003.htm>