Robin Lee Powell wrote:
On Mon, Nov 13, 2006 at 04:24:49PM -0500, Bob LeChevalier wrote:But if people were regularly running their text through checkers that caught such errors, I indeed think that people would stop making the errors. Human beings don't parse like machines, and thus accept things that a machine might not find legal. That just means that we need machines to help teach us to follow some of the rules.1. The official parser *does* accept {la.stivn.laitl.} as la followed by two cmene. So if you're right that all we need is computer training, we've already been having it for some time now, and it agrees with the proposed rule and not current rule.
That's what I said - the official parser had minimal sophistication for lexing.
2. When you've written such a tool, let me know. Right now, you have no evidence that this is a human learnable rule; you're just having a pipe dream.
Of course it is a human-learnable rule. You underestimate human capabilities.
The issue seems to be whether it is a rule that humans will learn to follow *without error*, and at this point I would argue that we aren't sure that humans can learn when elided cu and elided terminators are valid without error. Humans don't detect those errors too easily either, and they become important in the longer and more complex sentences that a "native" Lojban speaker will likely use.
lojbab To unsubscribe from this list, send mail to lojban-list-request@lojban.org with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if you're really stuck, send mail to secretary@lojban.org for help.