The best bet I've found for reading the contents of a PDF doc when it's displayed on screen is implementing the MSAA features of TestComplete. There are a number of posts up here on that topic so, rather than rehashing them, I'd suggest a quick scan through to see what you can find.
If, however, you are looking to verify the contents of a PDF without actually opening it in a reader (like reading the contents of a CSV file), that I don't know.
I'm still looking for a way to get the rest of the text. The words by itself is pretty limiting to the amount of verifications I can do. Ideally, i'd like to convert it to xml or html so I can examine table structures etc... so I can corrolate data points with their appropriate labels.