What is in a link: using the web to find semantic relations

Googling around I was trying to find a connection between empirical approaches for generating ontologies and I stumbled across one of my old colleagues of the Media Lab community: Carson Reynolds. Recently, he wrote a small prototype that was googleing for a word and then parsing the html pages of the main links obtained by the search engine. Then taking a list of English words with emotional orientation he was trying to mach these words with the boo-horay theory stated by the Cambridge philosoper A. J. Ayer.

An example query: emotemail

Boo Hooray
critical 21 content 102
bored 9 happy 30
confused 7 interested 28
surprised 4 ready 19
angry 3 patient 9
disappointed 3 satisfied 2
rejected 3 accepting 1
annoyed 2 agreeable 1
bitter 2 calm 1
contrary 2 pleased 1
irritated 2 sympathetic 1
suspicious 2
ambivalent 1
awed 1
disgusted 1
jealous 1
uncertain 1

This simple concept reminded me of the paper presented by Massimo Poesio during the FOIS conference. He was using the internet as well for finding semantic relations. Therefore I checked his references and I found that is not a very new approach. In fact, many people tried already to use the net as super-referent to define semantic relations. Some of the links I have to check:

1_ G. Grefenstette. SEXTANT: Extracting semantics from Raw Text. Heuristics, 1993.
2_ F. Keller, M. Lapata. Using the Web to Obtain Frequencies for Unseen Bigrams. Computational Linguistics, 29(3), 2003.
3_ M. Poesio and Abdulrahman Almuhareb, 2004. “Feature-Based vs. Property-based KR: An Empirical perspective”, Proc. of FOIS, Torino, October.

Leave a Reply