[Xapian-devel] GSOC 2014

[Olly Betts]

On Tue, Feb 25, 2014 at 01:19:23PM -0800, Jiarong Wei wrote:
> From the list of project’s ideas, Learning to Rank interests me a lot.
> Last year, I did a course on Information Retrieval at my university.
> Also I’m interested in Machine Learning. So, that’s why letor that
> combines both techs attracts me. I’d like to know more about it by
> improving the code. I also notice the potential mentor Parth Gupta is
> also the initial contributor of letor as GSOC 2011 student. That's so
> cool!

Yeah.

Parth knows most about the details of the Learning to Rank code, as he
either wrote it or mentored the student writing it.

But in more general terms, a letor project this year needs to get the
code to a point where people can actually use it easily.  The letor
module ideally needs to be trained for a particular application, so we
need to ensure that is something that's feasible for a typical user to
do.  If it's too incomplete or too buggy or too hard to set up, it's not
going to get used, and that would be a real shame.

It's OK to add new features too, but the focus should be on making it
all work well, sorting out any new xapian-core features which would help
the letor module, and getting decent test coverage of the letor module.
Some sort of "HOW TO" document describing setting it up would be good
too.

> Now, I’m reading the resources provided on ideas’ page. Do you have
> any other suggestions of knowing more about the letor?
> And I’d like to test the function of letor. But I can’t find code
> example. Can u give me some suggestions?

Hopefully Parth can help here.

Cheers,
    Olly