Image via [Wikipedia](http://commons.wikipedia.org/wiki/Image:Fivestagespipeline.png)
I tested on a directory containing mixed bag of PDFs, books, book chapters, articles and misc stuff. Most of the time the data was extracted but I ended with a wrong reference or garbage information about the file. Not mentoning that it took almost one full hour to analyse a little bit more than 500 files. I think I need to buy an eight-core machine now. The CPU peaked at 25-30% (one full core for it) on Vista, and the memory got up to 300 Mbs.
Not impressed, maybe one day.