Why Wikipedia Needs to Make Friends with WordNet

Data of various kinds acquired from Wikipedia is gaining popularity in NLP and related areas of research. For one reason, Wikipedia provides us with data with broad coverage. Its coverage is so broad that no other freely available linguistic resource can match it. It is often claimed that this is evidence for the triumph of “collective intelligence.” Radical enthusiasts of Wikipedia even go on to claim that researchers in NLP and SemanticWeb no longer need WordNet (WN) (Fellbaum, 1998).1) They allude to the superiority of Wikipedia-derived data over manually crafted data like WN in terms of development ease, speed, and cost as well as coverage. WN comes with precision endorsed by psychological reality that most WWWderived data lacks, but some people also tend to criticize the subjective nature of the word senses that WN specify, no matter how finegrained its sense distinctions are. All in all, they seem to try to dismiss WN-like lexical resources by suggesting that they are outdated in the age of WWW. And here comes the crucial question, Does Wikipedia dispense with the need for WordNet? In this paper, we argue that the answer is