You need to have enough documents that are similar to each other in a given area to train the topic assistant. This is not an easy task, IMHO and can be rather fickle. Try getting between 15-20 documents in the same doc lib in an area, then train the topic assistant. See if that helps. Since I’ll be doing this in my advanced class in May, I’d appreciate any real-world feedback and experience you could toss my way.