Newsgroups: rec.arts.int-fiction
Path: gmd.de!xlink.net!news.belwue.de!news.dfn.de!scsing.switch.ch!swidir.switch.ch!univ-lyon1.fr!jussieu.fr!math.ohio-state.edu!cs.utexas.edu!uunet!world!dmorin
From: dmorin@world.std.com (Duane D Morin)
Subject: Re: Parsers
Message-ID: <CM20Bz.74D@world.std.com>
Organization: The World Public Access UNIX, Brookline, MA
References: <1994Mar1.140641.23498@schunix.dmc.com> <2l2jq2$17l6@msuinfo.cl.msu.edu>
Date: Wed, 2 Mar 1994 20:16:47 GMT
Lines: 43

In article <2l2jq2$17l6@msuinfo.cl.msu.edu>,
Steve Dunham <dunham@cl-next4.cl.msu.edu> wrote:

[on the subject of ambiguity and parser interpretation]

> 1. If there are no semantic interpretations consistent with the game
>world, then complain to the user (``I don't understand'' or ``there is
>no box on the table'').

Easy enough.

> 2. If there is one interpretation, act on it.

Again, goes without saying.

> 3. If there is more than one interpretation, any of the following
>actions could be taken:
>
>	(iii) Give a `unlikeliness' score to each of the possible
>	semantic interpretations (mark unusual interpretations or
>	rarely used constructs as being more unlikely, add them up for
>	each interpretation), and if any one interpretation stands out
>	as being much more likely than the others, act on it.

Fuzzy?  Wow.  Problem with it is that, to be robust, you'd still have
to verify with the user that you'd guessed correctly.  Well, you
wouldn't HAVE to, but in my day job I write medical software and 
there's no guessing allowed, everything has to be verified by the user. :(

>unlock the door
[unlocking the big door]

...

>unlock the door
You mean the big door?

>Steve Dunham
>dunham@gdl.msu.edu

Duane


