[Project-ideas] Improving IR methods for OCR of Indic Scripts

[Sankarshan Mukhopadhyay]

On Wed, Apr 24, 2013 at 1:04 AM, ABHISHEK GUPTA <abhi.bansal21 at gmail.com> wrote:
> I am a 3rd tear student at Dhirubhai Ambani Institute of Information &
> Communication Technology. I am interested in doing some work with
> Ankur-India on the topic "Improving information retrieval methods for OCR
> data sets consisting of Indic scripts". I want to know more about the
> project. What is the project's current state. What corpora, tools,
> algorithms and approaches are you using. As project is aiming at improvement
> of the method, what are the current results?

The idea of the project is to work on an upstream centric method which
will enable information retrieval with greater accuracy. The list
archive has a couple of threads on the OCR and IR related topic,
please have a quick read through them.

To answer your query on the "state of the union", there exist a
variety of approaches upstream in a divergent manner. The focus of our
organization's GSoC is to try as much as possible to extend and
enhance existing projects. With regards to the viability of current
methods of retrieval, please read up literature available. There have
been a few recent papers published from IIT-KGP among other places.


--
sankarshan mukhopadhyay
<https://twitter.com/#!/sankarshan>