Princeton’s WordNet and PHP

April 11th, 2006 | Tags:

Wordnet is Princeton’s online lexical reference system. It’s a wonderful piece of work really. Wordnet prints out synonyms, meanings and what not of words depending on the use of the word. (That’s the best way I can describe it in layman’s terms.) It also has the power to differentiate between nouns, verbs, adjectives and adverbs and the data of a word is stored depending on the context in which the word is used.

An online interface written in perl can be found here. Searching for the word “see” returns results where “see” is used as a noun and a verb. Pretty.

I had a small project where I had to import all the wordnet data into a MySQL database and then list out all possible synonyms of a word ignoring the context. Not too difficult right? Import the database, search for the word, look in the synonym table and you’re done? Wrong.

Let’s start with the importing into MySQL. There is no lack of contributors to the project. There is even a contribution of MySQL tables but those were version 2.0 [link]. I needed the latest version so all that was left to parse the prolog files on my own. The program is anything but simple. And that’s in no way a bad thing, it just means that my grasp of the english language isn’t all that great. I couldn’t figure out what file to use for what. I didn’t want to parse them all and was trying to figure out the absolutely essential files. Took me an hour or so but I ultimately got it. The wordnet forums were of absolutely no help. Yeah, I understand that this application wasn’t built to just pull synonyms and stuff but that doesn’t mean that it shouldn’t. Everyone on the forums expects you to understand every single details about wordnet and then, “you’ll get it”. Well no thanks.

Back to parsing the files, I somehow managed to parse the prolog files into a MySQL table, wasn’t that difficult once I figured out the files I needed (which was just the synset pointers files). Then came the part about checking for the words. Believe it or not it was more difficult than I expected. Again, it took me bit to figure out the wuick and dirty SQL which would be :

select s2.word from wn_synset s1, wn_synset s2 where s1.synset_id=s2.synset_id and s1.word=”talk” group by s2.word;

Now I didn’t do it like that because I had to list out the word context as well but I picked up that one liner SQL from here.

That said and done I really like what princeton has done with wordnet. I’m planning to build an online PHP solution based on wordnet sometime soon. (As soon as I find some free time). In the meantime i’ll be putting up parsed wordnet SQL files soon to help those in urgent need.