05 November 2007

Literal Semantic Web

I was thinking recently how I would like to understand the concept of the semantic web a lot better. Well, now I have found one I like better, in the form of Visuwords, a self-proclaimed online graphical dictionary. It doesn't seem to be functioning very well at the moment (I keep getting the error message "XML Load Failed") but I can get some idea.

Enter exquisite (bexquisite not yet having entered most traditional dictionaries) and you get a visualisation of the network around the word - around the central node, you get a number of nodes each representing one sense of the word. Spanning out from those, you get synonyms (or close synonyms, full synonyms being rare in natural language). For example, sharing the definition "of extreme beauty" you have the word beautiful. Of course, the two words have many different connotations and hence are not full synonyms but they share many of the same necessary and sufficient or prototypical qualities. You also get the related noun exquisiteness (though not bexquisite).

This example is quite simple and produces a small network of definitions but the site isn't working well enough for me to test much more extensively (probably because whichever blog on which I found the link has sent a whole load of traffic bouncing their way).

I always liked the Connectionist model of the representation of language in the mind/brain. This is perhaps because I particularly enjoyed the speech error component of my psycholinguistics paper, where we used different types of speech error (such as malapropisms or Spoonerisms, to name a few of the best known categories, but also more commonly, sound substitutions or deletions or repetitions) to support different theories of language. As a pedant and a sharp-tongued one at that, I was always particularly quick to pounce on speech errors made by friends and in an effort to look creative in my exams, I used to carry around with me a notebook where I would record interesting and real speech errors my friends made.

In any case, I always liked those errors where more than one component of language was involved (e.g. phonetics and semantics (the sound and the meaning)). For example, in common speech errors involving sound deletions, people tend to produce actual words (or phonologically possible words of a language) more often than chance would suggest (hence bed and butter is more a likely error than bread and brutter, because of this real-word bias). Similarly, I liked the Connectionists' networks of words, which were linked to one another by semantic, phonetic and syntactic properties and activating a word such as good, could cause many other words to be activated (could (phonetic similarity); better (morphological similarity); nice (semantic similarity); orthographical (goon); etc.).

I don't know whether Visuwords has the contents of its dictionary linked in so many ways; I suspect they are only linked semantically. However, it is still interesting for a linguist and I suspect I shall be browsing their dictionary much more once the traffic has died down. To be fair, though, I used to read the dictionary when I was ten so I could learn new spellings and definitions (hence my knowledge of orthography and semantics greatly outstripping my knowledge of pronunciation) and was perfectly happy browsing through even though I could only do so alphabetically. I guess I'm just a word geek. A verbiphile. A weirdo.

No comments:

Post a Comment