
www.Usenet.com
| <-- __Chronological__ --> | <-- __Thread__ --> |
Matthew Purver <[EMAIL PROTECTED]> writes: > Rod Davison wrote: > > > On Thu, 06 Nov 2003 09:24:17 +0100, Niki Estner wrote: > > > > Question: Why is NLP a hard topic? > > Answer: Because natural language is a a rules based system where users > > don't follow the rules. > > > > Seems odd, but you have to understand that language is a social phenomena. > > well, no, according to my rules, it's a social phenomenON. But as you so > rightly point out, users don't follow the rules :-D > > actually, I think that's quite a good example: I could understand what you > meant no problem at all of course, but a machine programmed with strict > plural morphology rules would have problems. Exactly. Which is one reason I think it's a shame that NL parsing is so focused on not overgenerating. Almost the first thing most people do when developing a natural language grammar is to lock down agreement, so that it's impossible to generate or parse "He eat apples". But even a child can understand "He eat apples". I can understand why this path is tempting. Case agreement is one of the first pieces of prescriptive grammar most people learn, and handling case agreement immediately gives a grammar a superficial Eliza-like appearance of polished correctness. Nevertheless, I think grammar proper is the wrong level to handle agreement on, except for purely prescriptive agreement. If I had to give the NLU community one piece of advice, it would be to loosen up about grammar overgenerating. -- Tom Breton at panix.com, username tehom. http://www.panix.com/~tehom
| <-- __Chronological__ --> | <-- __Thread__ --> |