Hi,
I am happy to inform, that Shrinivasan has created a python script to automate the process in Linux system. This scripts upload the PDF files to Google Drive, download the OCRed text and split, merge the text files properly to fit as the PDF file. We have just tested the script for small files in Kannad and Bengali Wikisource and it was successful. We are going to test the script for using different types and sizes of files and in other Indic languages in next few days.
The script is in https://github.com/tshrinivasan/OCR4wikisource
Regards, Bodhisattwa
On 2 December 2015 at 17:21, Tobias Schönberg tobias47n9e@gmail.com wrote:
I think it is important for non-technical readers of this list to separate the 2 issues in the discussion.
- OCR-Integration
This is something WMF can help with, because they can make the connection between an OCR service and Mediawiki easier and automate certain steps.
- OCR
WMF is not programming an OCR-software and it would probably be a bad idea to reinvent the wheel. It would be far better if editors reached out to existing ORC-software projects. Starting a discussion or filing a bug is an important first step in improving the situation. Tesseract-OCR (https://github.com/tesseract-ocr) for example is an open-source project that works on OCR (No bugs filed for e.g. Bengali). The mailing list (https://groups.google.com/forum/#!forum/tesseract-ocr) contains discussions about e.g. Bengali ( https://groups.google.com/forum/#!searchin/tesseract-ocr/Bengali). So I think the situation might not be good, but is certainly on its way of getting better. Maybe WMF-India can fund a developer to work on Tesseract-OCR. Another idea would be, to reach out to local universities. Maybe a few informatics-students can improve the situation.
-Tobias
2015-12-01 19:51 GMT+01:00 ViswaPrabha (വിശ്വപ്രഭ) viswaprabha@gmail.com :
From that page which, Alex has linked: "On the other hand, using the service for converting document formats *is* SaaSS, because it's something you could have done by running a suitable program (free, one hopes) in your own computer."
Hundreds among us have burnt their hands in developing a successful 'free' OCR tool for Indic languages without any real luck until now. Until such a tool appears on the horizon, the Google facility is just okay to be used.
Especially so, because we are anyway dealing with 'free' input and output material.
-Viswaprabha
On 1 December 2015 at 21:49, Bodhisattwa Mandal < bodhisattwa.rgkmc@gmail.com> wrote:
Hi Alex,
Of course, building free OCR can be the only permanent solution, but WMF is not interested in building new OCR right now. The language engineering team said at the conference that, they don't have the infrastructure and expertise to build such software. That's why, we have to rely on Google OCR, knowing very well about its profit making intentions. It's just a temporary solution but right now, its the only best possible alternative for us.
Regards Bodhisattwa On 1 Dec 2015 21:12, "Alex Brollo" alex.brollo@gmail.com wrote:
... nevertheless I found very interesting this about "SaaSS": https://www.gnu.org/philosophy/who-does-that-server-really-serve.html
So, to build a true, excellent and indipendent "wikisource multilingual OCR service" would be a better solution.
Alex
2015-12-01 16:06 GMT+01:00 Bodhisattwa Mandal < bodhisattwa.rgkmc@gmail.com>:
Hi Nemo,
Thanks for your interest. You can find the list of Google OCR supported languages in the following link -
https://support.google.com/drive/answer/176692?hl=en
Regards, Bodhisattwa Thanks for posting about the topic. Which indic languages are we talking about exactly? Are they included in the recent FineReader versions now used by Internet Archive?
Nemo
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l
Wikisource-l mailing list Wikisource-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikisource-l