Saturday, June 7, 2008

Error-correcting language

How large would our language have to be to be self-correcting, or at least error-flagging? If you're at all familiar with error detection schemes, you know that the goal is to present the receiver of a message with: 1) the information needed to correct a faulty message, or 2) minimally, the information to know that the message is faulty, allowing a request to retransmit.

What I want is to be able to type a passage in which I would have all typing errors flagged. For example, I often type "tome" instead of "time"; since "tome" is a legal word, it isn't flagged by spell check. So I guess what my real question is, how could we create a language in a specific keyboard context for which we couldn't have those errors?

But then, I also wouldn't want the kind of errors where I forget one keystroke ("wan" is a word, so leaving the final 't' off "want," as I just did, should be flagged). That makes creating the language even more difficult ("an" could be a word, but we couldn't then have "any," "ant," etc.).

This strikes me as a worthwhile problem for someone to work on, if it hasn't been done already (not that I'm volunteering). I suppose any such language would feature too many long words, would probably not be utterable, would be unworkable. It is kind of an interesting problem, though....

No comments:

Clicky Web Analytics