[MUD-Dev] Text Parsing

Travis S. Casey efindel at io.com
Wed Jun 2 10:47:50 CEST 1999


On Wed, 2 Jun 1999, Katrina McClelan wrote:

> a huge hoard of orcs spot you and reach for their weapons with a snarl.
> -> take off running away from the orcs.
>
> -> put on cloak.
> you put on the white cloak.
> -> look around.
> an off white cloak lies on the ground here.
> -> take off white cloak.
> 
> (If you have a parser that will not blarf all over any of these, I'm
> _much_ impressed)

Depends on what you mean by "blarf".  My own parser would have done the
intended thing for all but "take off running away from the orcs."  For
that, it would have given back an error/help message, like:

Usage:  take <object or object list>
Use the command "help input" to learn more about how to make the mud
understand your commands.

> worse, when you throw in the tendency for experienced players to perfer
> lazy typing to english, you get problems:
> 
> the lizardmen have left an offering to their gods on the alter.
> -> take off alt
> (shorthand for take offering from alter that most mud processors allow,
> and that veteran players will expect to have work)

Most Dikus and Diku-derivatives, you mean.  In over eight years of
mudding, I never saw this sort of thing until after I joined this list --
because I played LP and Tiny derivatives.

(If I saw that on a mud which was supposed to have "natural language
input", the first thing I'd do is complain -- it's "altar" not "alter".
Remember -- output should be more correct than input is required to be.)

> An English parser would be neat, but there are a lot of problems
> associated with it... especially since as Ben pointed out, most players
> beyond newbiehood are going to type the shortest amount possible to get
> the desired result.  I hate to be a cynic, but alas it's my nature:  I'm

Parsing a greater subset of English than most muds do is not mutually
exclusive with allowing players to type the shortest amount possible.

> not entirely sure that such a thing is feasible for a mud, unless you're
> either willing to accept some strangeness in it, or to restrict the
> sentence structure some (force the active verb to be the first word for
> example).  At the very least I don't think anyone would argue with me

No one who has been contributing to the discussion of parsing has
suggested making a parser that *doesn't* require the verb to be first.

> saying that at the very least this is a very large and ugly problem. The
> discussion is certainly interesting though, and of course, people told the
> Wright brothers that they were nuts to be trying to build a plane too,
> until they managed to do it.

I think this is becoming a strawman discussion.  To reiterate what I've
said in this message and a couple of others:

 - There's a long way between trying to improve mud parsers by making use
   of NLP ideas and trying to implement full "natural language" input.

 - Allowing input that's closer to English does not require *not* allowing
   traditional mud-style commands.

 - None of the people who have been giving suggestions on how parsers 
   raised the term "natural language", and none of them have suggested
   making traditional mud commands *not* work.

--
       |\      _,,,---,,_        Travis S. Casey  <efindel at io.com>
 ZZzz  /,`.-'`'    -.  ;-;;,_   No one agrees with me.  Not even me.
      |,4-  ) )-,_..;\ (  `'-'
     '---''(_/--'  `-'\_) 



_______________________________________________
MUD-Dev maillist  -  MUD-Dev at kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev




More information about the mud-dev-archive mailing list