Re: [xml-dev] Structured from/within unstructured documents

["Stephen Green" <stephengreenubl@gmail.com>]

I notice there are commercial tools advertised to convert PDF to .doc
or .odt, etc
or to extract data in one way or another. How reliable do people find
such tools?
Is it realistic yet to be extracting data and converting it to, say,
XML documents
in large volumes and with crucial data such a financial, technical or medical
records?

On 16/12/2007, Edward C. Zimmermann <edz@bsn.com> wrote:
> On Sun, 16 Dec 2007 18:15:05 +1100, Greg Hunt wrote
> > Stephen,
> > The problem with processing the physical PDF file is precisely its
> presentation orientation.
>
> You have to render PDF (at least internally into a buffer). Its a format with
> graphical "language" not totally unlike (and built-upon) PostScript where,
> among a host of features, each individual character can be positioned.
>
> Popular "freely" available PDF tools that can be used to "extract text"
> are, among others, Adobe's Acrobat Reader, Derek Noonburg's Xpdf, Poppler
> and Ghostscript. M$ Windows includes a "filter mechanism" called iFilter
> for their own search. It includes apparently, among others, a filter
> supplied by Adobe intended for the extraction of text from PDF.
>
> > A perverse document can mix image and text or even embed the text in the
> reverse order that it would be displayed in.
>
> Not really that wholly uncommon--- calculating glymph position from the right.
>
> >
>
> In rendering, however, you need or want to keep paragraph blocks together
> and **not** (as the case from a "screen scrape" of a display rendered page)
> preserve the columns and visual flow elements as these not only make it
> much more difficult to extract simple things like sentences but also
> don't deliver any contextual information. That a text was set in two column
> with a center picture is a result of its chosen style and not content
> structure--- recall that different output devices can look different.
> Linking structural semantics for many of these style elements is tenuous.
>
>
>
> --
>
>  Edward C. Zimmermann, Basis Systeme netzwerk, Munich
>  Office Leo (R&D):
>    Leopoldstrasse 53-55, D-80802 Munich,
>    Federal Republic of Germany
>  http://www.nonmonotonic.net
>
>


-- 
Stephen Green

Partner
SystML, http://www.systml.co.uk
Tel: +44 (0) 117 9541606

http://www.biblegateway.com/passage/?search=matthew+22:37 .. and voice