[Project-ideas] Improving information retrieval methods for OCR data sets consisting of Indic scripts

[Sankarshan Mukhopadhyay]

Hi Rabindra,

Thank you for writing in.

I am replying as a top-post because I have copied in the mailing list
we use to discuss project ideas (subscription interface should be
available from <http://lists.ankur.org.in/listinfo.cgi/project-ideas-ankur.org.in>

I have also added Jayanta Nath in the list. I met Jayanta yesterday
(after a suitably long period of interactions over email) and, we
ended up chatting about the usual - "how to crack this OCR issue in a
manner that helps the Bengali Wikipedia community and, especially
Wikisource"

I am glad to note that you have taken a look at Abhishek's existing
work. Have you been able to reach out to him and discuss in some level
of detail the current state of the work? The voting piece is somewhat
based on the concept that a larger number of users of the system can
help train the system for higher degree of accuracy.

ankur.org.in will be putting in an application as a mentoring
organization. However, the acceptance in GSoC2014 is always subject to
- [1] good set of project ideas; [2] reasonable success from previous
year etc. So, there is a period of waiting before one gets to know
about being selected as a mentoring organization and, thereafter
begins the process of selecting strong applications from students.

I would recommend that you spend this time catching up with Abhishek
and also Jayanta in order to be able to understand a real-life
utilization of your project (should ankur.org.in be selected and, you
are accepted as a student)

/sankarshan

On Mon, Feb 3, 2014 at 12:56 PM, Rabindra Rakshit <rovir2r at gmail.com> wrote:
> I (Rabindra Rakshit), am interested in applying for GSOC 2014, and would
> like to know if Ankur India is applying as a mentoring organization this
> year also.
>
> I am currently pursuing my B.tech in Computer Science(CSE) from College of
> Engineering and Management, Kolaghat, and being born a Bengali, would love
> to see my language flourish in the open source community.
>
> I am particularly interested in the project about Improving information
> retrieval methods for OCR data sets consisting of Indic scripts(Info
> Rescue). I had a look on the work plan of Abhishek Gupta, the final voting
> system in a general(abstract) manner is yet to be implemented.
>
> I don't have any exact experience about OCR, but I do have experience of
> working with Information Retrieval Systems, in fact, right now I am working
> on Consensus Sequence Segmentation, an Unsupervised Text Segmentation
> algorithm that relies entirely on statistical relationships among alphabets
> in the input sequence to detect location of word boundaries. I have attached
> a document of our work which is still in progress.
>
> Link: http://arxiv.org/abs/1308.3839

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