Path: utzoo!attcan!uunet!lll-winken!csd4.milw.wisc.edu!leah!bingvaxu!sunybcs!sher
From: sher@sunybcs.uucp (David Sher)
Newsgroups: comp.ai
Subject: Re: Question on Chinese Room Argument
Keywords: Symbolic Systems
Message-ID: <4440@cs.Buffalo.EDU>
Date: 1 Mar 89 19:04:21 GMT
References: <45126@linus.UUCP> <5662@homxc.UUCP> <45199@linus.UUCP> <917@jhunix.HCF.JHU.EDU>
Sender: nobody@cs.Buffalo.EDU
Reply-To: sher@wolf.UUCP (David Sher)
Organization: SUNY/Buffalo Computer Science
Lines: 19

Just to test out what is and isn't a symbolic system.   Consider a
stochastic context free grammar.  This is a grammar that has
a probability associated with each production.  Thus each element of the 
language it accepts has a probability associated with it (along with
a certain probability that the machine never outputs anything).  
Now consider a machine that takes a stochastic grammar and an input string
and outputs the most probable parse tree for the input.  Is this machine
doing symbolic processing?  Or is more information required to answer this
question.  Assume it used a modified form of Earley's algorithm to do
this.  Now is it doing symbolic processing?  

If enough people are interested in how to modify Earley's algorithm to 
accept stochastic grammars I can post (or even write a paper on the topic
if it hasn't been done yet).  Its fairly trivial.


-David Sher
ARPA: sher@cs.buffalo.edu	BITNET: sher@sunybcs
UUCP: {rutgers,ames,boulder,decvax}!sunybcs!sher