GSOC Proposal

Olly Betts olly at survex.com
Mon Mar 26 05:27:22 BST 2018


On Wed, Mar 21, 2018 at 11:48:03PM +0530, Mehak Goyal wrote:
> My understanding of the project is that we are trying to train LeTor
> through User clicks without the user having to train explicitly with
> relevance judgment. To capture the relevance judgments, currently, we have
> deployed the counting algorithm based DBN model and we wish to further
> improve upon it using Expectation Maximization and Forward-Backward
> algorithms and also include other models (Dependent Click Model and Intent
> Aware Model) for comparison to the user.

That sounds about right.

> I understand I have a considerable grasp of the concepts and I wish to know
> how to proceed further to be able to draft the proposal. What is expected
> of me at this stage when I am done with building Xapian and reviewing the
> available, relevant literature.

You need to work out how you're actually going to go about implementing
this in Xapian.  Not every last detail needs to be planned out, but you
need to analyse and investigate enough that we can all be confident that
what your proposing is feasible for you to implement in the 3 month
timescale of GSoC.

There are some tips and pointers to useful resource to help with this in
comments in our proposal template, but you're welcome to discuss this
here or on IRC.

> I am very much interested in working with Xapian given my background and my
> interest. A prompt response shall be appreciated.

Sorry nobody replied sooner.  I didn't reply initially as I was hoping
Vivek would as he knows more about this project idea (he worked on this
as a student last year), and then it was buried by newer mails.

Cheers,
    Olly



More information about the Xapian-devel mailing list