Friday, November 26, 2010

Wordnet is really cool

Here's something that happens to me all the time: I'm writing a story, and my hero is in, perhaps, a forest. And he's looking around the forest, and he sees some trees. And this where I get stuck, because I can't for the life of me figure out what kind of trees he's looking at.

We know this: specificity is good in writing. "Bob looked at the tree," is not nearly as potent an image as "Bob looked at the oak." With the substitution of a single word the image goes from vague (and thus unlikely to excite the interest of the reader) to much more concrete.

The next time I find myself stuck for specificity, I'm heading to Wordnet. But what is Wordnet? Wordnet is really cool.

At first glance, Wordnet appears to be a dictionary. Type in a word like tortilla and you get a definition:

S: (n) tortilla (thin unleavened pancake made from cornmeal or wheat flour)

but the curious among you might notice that S: there in front of the definition. That S: is the key to Wordnet's really coolness.

You see, that "S" stands for "semantic" and semantic relationships are what Wordnet is all about. Beyond giving the meaning of words, Wordnet connects those meanings to each other through the kind of organizational architecture that would make a semiotician drool. If I click on the S:, here's what I get:

Ok then. That's a lot of funny looking words. What do they mean? Well, hypo- and hypernyms are the easiest to explain. As you will remember from your intimate knowledge of Greek, "hypo" means "under" (as in "hypothermia" or "hypodermic") and "hyper" means (above, as in "hyperactive").

Tortilla is a pretty specific word. But its hyponyms are all the words that are even more specific than tortilla. In other words, if you wanted to find out what different types of tortilla there are, you would click on "direct hyponym." Let's take a look-see:

A-ha! A tostada is a kind of tortilla. The next time I'm writing a scene in a mexican restaurant and I want to give it some real flavor, I'm going to drop a mention of a tostada!

Hypernyms go in the other direction. Let's say I wanted to get a little more general. I would click on "direct hypernym" and:

 So there you go. A tortilla is a kind of pancake and a kind of hot-cake etc. I think that's pretty cool because I never would have considered a tortilla to be a kind of pancake, but now that I think about it, it is.

Ok, so what about "inherited hypernym"? That's even cooler. When you click on that link, it'll get more and more general until it has reached the most general word it knows:

A tortilla is a pancake. A pancake is a cake. A cake is a baked good. A baked good is food. Foods are solids. Solids are matter. Matter is a physical entity. A physical entity is an entity. And as far as Wordnet goes, there's nothing more general than an entity. It's at this point that Wordnet becomes more than just a crazy super-dictionary and becomes more like a philosophical argument.

How about sister terms? Sister terms are all the words which are exactly as specific as the word you've selected. So:

Here you have a list of all the different kinds of pancake. Pretty cool. If you want to get some more local flavor in your story, don't have your Russian soldier eat a pancake. Have him eat a blini (unless he's on vacation in Germany. Then he could eat a pfannkuchen).

As I hope you can see, Wordnet can be an invaluable resource to a writer, not to mention a good way to spend a couple of hours if you just happen to be a lover of words. So how about my original question. My hero (let's call him Bob), is wandering through the woods, lost and starving. He sees a tree:

S: (n) tree (a tall perennial woody plant having a main trunk and branches forming a distinct elevated crown; includes both gymnosperms and angiosperms)

Ok. What kind of semantic goodness can Wordnet hook me up with?

AWESOME!!. Bob is saved, because that's not just any tree. It's a quandong,

I've only just scratched the surface of what Wordnet can do. There are meronyms and holonyms and frequency counts, oh my!

By the way, Wordnet is entirely created and maintained by human beings. Over twenty years in the making, but I bet there are still places where you can think of semantic relationships that haven't been entered yet. That alone provides a fun challenge for the vocabulary-endowed. Enjoy!


  1. I wonder if it yet possible to explore meanings for which there are as of yet no words. Seems like a pretty tall order, even for glorious Wordnet, but if there are such complex associations between words, could there not also be complex associations between meanings and therefore an ability to see so-called semantic gaps?

  2. That would be really awesome and potentially lucrative if you were interested in cornering the market on creating words for concepts that don't have words for them yet. Kind of an advanced version of the interrobang or The Meaning of Liff.

    I don't think there's exactly a way to do that yet. I suppose what you would be looking for is to see a root word being modified by the same descriptors. For instance, if you see "status" and "update" together a lot you might be able to identify a need for a word like "tweet" to fill the semantic gap. This is a poor example but I think the concept is sound -- you're looking for phrases that are on their way to becoming de facto definitions, as if you were using the dictionary backwards.

    Theoretically you could create a computer program to do something like that -- after all, in some ways it's just an advanced form of Google Trends, where instead of just looking at trending words, you're looking at trending phrases. That would only get you half way though because you'd still need a way of knowing which phrases describe things that already have words for them and which don't.

    I really love the idea of "semantic gaps." They're where language evolves. We know they're there but other than human ingenuity and guesswork we have no way of finding them (that I know of).

    Sidenote: in poking around Wordnet for this comment, I learned the words "Cimmeran," "caliginous" and "aphotic."

  3. I think it would really help me and other people who are learning English as their second language to express their thoughts clearly. I have a lot of trouble expressing the image in my head through English. It is really common that we can't find a correct word for the expressions, but with Wordnet, I think it would help lots! I will use Wordnet to write my MFP..!!

  4. What a wonderful tool! I always just added adjectives until things sounded cool.
    "Bob looked at the tree."
    "Bob looked at ancient tree."
    "Bob looked at the nearly dead ancient withering tree."
    "Bob looked at the awesomely tall, towering, decrepit and nearly dead ancient withering tree"
    but now i can say
    ""Bob looked at the awesomely tall, towering, decrepit and nearly dead ancient withering oak tree"
    oh happy day! lol