Eric or others, do you know of any utility that converts a PDF and retains coding for where font or font-style changes? Or converts a web page with associated CSS and notes where font-style and HTML text block stops and starts?  It seems that would be the starting point for recognizing citation entities.  I've seen websites for FreeCite http://freecite.library.brown.edu/ and Parscit http://aye.comp.nus.edu.sg/parsCit/ through web searches, but don't know how close they got to the Grail before becoming legend.

Cindy Harper

-----Original Message-----
From: Code for Libraries [mailto:[log in to unmask]] On Behalf Of Eric Lease Morgan
Sent: Thursday, June 18, 2015 1:04 PM
To: [log in to unmask]
Subject: Re: [CODE4LIB] Desiring Advice for Converting OCR Text into Metadata and/or a Database

On Jun 18, 2015, at 12:02 PM, Matt Sherman <[log in to unmask]> wrote:

> I am working with colleague on a side project which involves some 
> scanned bibliographies and making them more web searchable/sortable/browse-able.
> While I am quite familiar with the metadata and organization aspects 
> we need, but I am at a bit of a loss on how to automate the process of 
> putting the bibliography in a more structured format so that we can 
> avoid going through hundreds of pages by hand.  I am pretty sure 
> regular expressions are needed, but I have not had an instance where I 
> need to automate extracting data from one file type (PDF OCR or text 
> extracted to Word doc) and place it into another (either a database or 
> an XML file) with some enrichment.  I would appreciate any suggestions 
> for approaches or tools to look into.  Thanks for any help/thoughts people can give.


If I understand your question correctly, then you have two problems to address: 1) converting PDF, Word, etc. files into plain text, and 2) marking up the result (which is a bibliography) into structure data. Correct?

If so, then if your PDF documents have already been OCRed, or if you have other files, then you can probably feed them to TIKA to quickly and easily extract the underlying plain text. [1] I wrote a brain-dead shell script to run TIKA in server mode and then convert Word (.docx) files. [2]

When it comes to marking up the result into structured data, well, good luck. I think such an application is something Library Land sought for a long time. “Can you say Holy Grail?"

[1] Tika - https://tika.apache.org
[2] brain-dead script - https://gist.github.com/ericleasemorgan/c4e34ffad96c0221f1ff

—
Eric