Taking the day off from work

So I'm taking the day off from working. I've been writing in my journal, suggesting things to the LJ suggestion community and doing a lot of thinking. It's been kinda' fun getting to know the LJ team. They're fun if a little cynical. My post to the suggestion board has inspired me to think about graphical pattern matching and language a bit more. I s'pose I should sit down and actually study language and syntax and all that if I really want to do something with my interest. But it's also fun being ignorant and just mentioning things that I would like to see, whether they're plausible or not.

So, back to language!

How did I learn English? I didn't know it when I was born... The way I learned it was by LEARNING it.

1. To gain knowledge or information of; to ascertain by
inquiry, study, or investigation; to receive instruction
concerning; to fix in the mind; to acquire understanding
of, or skill; as, to learn the way; to learn a lesson; to
learn dancing; to learn to skate; to learn the violin; to
learn the truth about something. ``Learn to do well.''
--Is. i. 17.

So why shouldn't we be able to write a program to learn language just like you and I learned it? Computers aren't smart, but they learn well. So we should learn 'em to learn. What exactly did I do while I was listening to my parents babble when I was a kid? I didn't understand. What inspired me to want to know what they were saying? How did I communicate? Well, I heard lots of patterns.

There, that's important. Find patterns and discover what they all mean. evan said that my previous musings on language didn't take context into consideration. He gave me this example:


I've never really understood part-of-speech tagging, anyway:
"a toy car factory".  "a" is an article(/determiner) and "factory" a
noun, sure, but what are the other two?  Nouns?  Adjectives?


So we just need to get our little parsers to take everything into context. I think that's what I was saying here. Take a big story and take the time to fill out a form for each word. If all of the information in the story is kept, then it will be in context. I can't tell what parts of speech "a toy car factory" has in it, so I shouldn't expect a parser to be able to, either. So let's set our sights on what humans can achieve and get computers to do THAT before we expect them to do the impossible.