WordNet uses humans to build a database of how words are related (synonym,
is-a, etc.). Perhaps most importantly, NELL builds its database by having
computer automatically learn to build it with relatively little help from
humans compared to WordNet. While NELL could (and probably should) learn
the kind of relationships that WordNet captures, it's currently given an
ontology that focuses much more on how particular instances of things
referred to by words are related. For instance, it learns about
particular sports teams, athletes, coaches, who is on what team, who
coaches what team, what sport each team plays, etc. If you search for
"giants" in NELL, you'll get information about the NY Giants football
team. If you search for "giants" in WordNet, you'll get the sort of
things you would expect to find in a dicationary.
bki...@cs.cmu.edu