On Sat, Mar 26, 2011 at 04:16:02PM +0530, Abhishek Mahato
wrote:> Weigting schemes: I think instead of developing new schemes the
> existing BM25 could be improved by using the stochstic optimization or
> something else. I would be happy to work on develop a new scheme if
> thats a better option.
I wouldn't necessarily rule it out, but I think you'd need to convince
us that you have a realistic chance of producing better results than
known weighting schemes (such as the DfR ones).
> Java Bindings: The topic is pretty straight forward just a trivial
> question, which platform is the work expected to be in preferably,
> Windows/Linux, as the commands and header files will change
> accordingly right?
They should work on all platforms ideally. I'd recommend doing
your main development work on Linux, as we can better support you if
there are any problems.
> Lastly, Im a first timer to GSoC, so are there anything else I should
> know or do which might tip the scales in my favor??
http://www.booki.cc/gsocstudentguide/ has plenty of good advice, and
a couple of sample proposals in the appendices.
I think my top tip would be to take time to write a quality proposal,
and be prepared to respond promptly to any comments on it. In past
years I've seen plenty of proposals with little useful information,
and some promising proposals which got discounted due to lack of
response to feedback we've given.
Cheers,
Olly