Eric,

Very interesting.  I've have been working with some existing pdf 
utilities with a goal of automatically extracting the abstract from 
technical reports, articles and dissertations that are to be bulk 
uploaded to our institutional repository.   I tried two of our documents 
through your system and the first one worked great.
The second tech report I tried however generated this error message:

Software error:

No words from which to create a cloud - see add(...). at 
/usr/local/share/perl5/HTML/TagCloud/Centred.pm line 229.

For help, please send mail to the webmaster (root@localhost), giving 
this error message and the time and date of the error.


Although based on some subsequent messages where you mention tesseract 
maybe I misunderstood and your tool only handles pdfs that have already 
been OCR'ed which would explain why the second document (which only 
contains page images) fails.

-Bob Haschart


On 10/11/2013 11:16 AM, Eric Lease Morgan wrote:
> For a limited period of time I am making publicly available a Web-based program called PDF2TXT -- http://bit.ly/1bJRyh8
>
> PDF2TXT extracts the text from an OCRed PDF document and then does some rudimentary "distant reading" against the text in the form of word clouds, readability scores, concordance features, and "maps" (histograms) illustrating where terms appear in a text.
>
> Here is the idea behind the application:
>
>    1. In the Libraries I see people scanning, scanning, and
>       scanning. I suppose these people then go home and read the
>       document. They might even print it. These documents are long.
>       Moreover, I'll bet they have multiple documents.
>
>    2. Text mining requires digitized text, but PDF documents are
>       frequently full of formatting. At the same time, they often
>       have the text underneath. Our scanning software does OCR.
>
>    3. By extracting the text from PDF documents, I can facilitate
>       a different -- additional -- type of analysis against sets of
>       one or more documents. PDF2TXT is the first step in this
>       process.
>
> What is really cool is that PDF2TXT works for many of the articles downloadable from the Libraries's article indexes. Search an article index. Download a full text, PDF version of the article. Feed it to PDF2TXT. Get more out of your article.
>
> PDF2TXT currently has "creeping featuritis" -- meaning that it is growing in weird directions. Your feedback is more than welcome. (I know. The output is ugly.) Also, please be gentle with it because it does not process things the size of the Bible.
>
> --
> [cid:116F6092-2AB6-4E95-8199-25639542726A]
>
> Eric Lease Morgan
> Digital Initiatives Librarian
>
> University of Notre Dame
> Room 131, Hesburgh Libraries
> Notre Dame, IN 46556
> o: 574-631-8604
> e: [log in to unmask]<mailto:[log in to unmask]>
>
> [cid:8DBE3E66-AAD0-40A0-A626-745EEEA175E5]
>
>