No. From brief testing I found such automatic systems to be unreliable.
However, perhaps I could add tagging suggestions based on content extraction / keywords.
Seems like just looking at tf-idf would be useful. Could result in some false positive tags, but this shouldn't actually break the system, just perhaps add a little noise to the tag list.