Words, words, words

Polonius: What do you read, my lord?
Hamlet: Words, words, words.

My grandson's first word was "Boo!" (and he said it to me). Then he began naming things, such as birds (which he classifies either as "duck" or "owl"), road vehicles ("car") and spherical objects — from peas to oranges to the moon, he names them all "ball". Recently he's been acquiring far more vocabulary and has started to string words together into phrases, if not quite sentences. Having begun with nouns, he's now incorporating other parts of speech into his vocabulary: verbs, a few adjectives and prepositions. He can't yet manage personal pronouns or adverbs (other than "where?" -- e.g."where teddy?"), but he can already convey a good deal of information:

Dig, dig, uh-oh! (Alexander's commentary on the demolition of a building that he and his father observed.)
"Man-on-moon rock" [= rocket / lunar module]. (He was taken to the Science Museum last month.)
"Mummy wash ball." (Watching his mother pick a tomato and wash it under the tap, he picked another and held it up to her.)
"Big red bus." (He lives in London.)
"Dig, dig, dig! Yellow digger." (Playing in the sand pit.)
"Light on, light off."
"Bye-bye!"

To my daughter's delight, he even came out with "Love Mummy!" the other day.

I didn't think Alexander could manage anything other than the infinitive or present tense form of verbs, until the other day he surprised me after a hiatus on our Skype link: "A'nder did hit keys, gran'a gone." (i.e. I accidentally touched a critical spot on the computer keyboard that caused my grandmother's picture on the screen to disappear).

Interestingly enough, it's not only human babies who venture, nouns first, into language. It seems that computers take this route as well. My husband has been pointing my attention to the development of OpenCyc, "the world's largest and most complete general knowledge base and commonsense reasoning engine" that purports to contain "hundreds of thousands of terms, along with millions of assertions relating the terms to each other, forming an upper ontology whose domain is all of human consensus reality" rather like the links in the language processing part of our own brains, it seems. Apparently, the boffins involved with this project have for the last 24 years been trying to teach computers what we call "common sense", but in order to do that they have to give them an adequate word classification system from which to make the sensible sort of connections we humans make without much obvious effort. To give the example Chris showed me, "if you look at 'CEO' you'll see it has a subject type (organisation) and object type (person)." Therefore the computer understands that "an organisation's CEO is a person."

This web page gives you the opportunity to check the system for yourselves. I decide to give it a go and type in a few words. It knows what an "echidna" is, but it knows neither the word "mankind" nor the modern, politically correct alternative, "humankind". It doesn't know "me", either. All it shows for those nouns and that pronoun are "nothing available". Ditto, "quintessence".

For "romance", it gives: "romance [love] Has something to do with romantic love, like sending flowers to a significant other, or a couple embracing." Which leaves out an awful lot else.

I'll try it with a verb, I decide, and type "pitch", but it can only identify this word as a noun, the first example given being "baseball pitch".

How about adjectives? I try "nice" and "city of France" comes up! That's a good one.

It also adds a reference to something technical: "Nice, The-Program". This is one biassed ontological knowledge base.

Other parts of speech leave it floundering:

"whatever": nothing available
"unless": nothing available
"also": nothing available. It doesn't even know the word "then"!

My husband, reading this over my shoulder, says that my comments are inappropriate, because what I'm trying to describe is an "upper ontology," not a dictionary. I'm still dismayed though, for I find that the Reasoning Engine's understanding of "us" is limited to "United States" and if I look at the entries listed under the word "forward", I find that they're either computer related or applicable to a military context. The potential for a misunderstanding in that context is no longer funny. It worries me considerably.