I am attempting to create some specialized AI/Expert machine with respect to processing paragraphs and making conclusions from that. Our company has a large need to process paragraphs to determine catagories based on the paragraph. The text that we are looking at is different from source to source, so there are many different ways that a paragraph can be written where the conclusion is exactly the same. Any leads would be great...Any sample/tool programs would be excellent...Flow Charts The ideal situation would be the machine learning from examples, adding wieght to certain string combinations on it's own... and such.

Check out the work of Ingrid Zuckerman, in particular NAG.

You should also do some research into Latent Semantic Analysis (LSA). A possible LSA solution to your problem might be to:

1) Categorize by hand a large number of different texts into your categories
2) Create an LSA database for each category
3) Assign new texts to the categories by how close their LSA score is to the categories' averages.