From - Mon Apr 19 11:18:17 2004 X-Mozilla-Status: 0001 X-Mozilla-Status2: 00000000 Path: acsu.buffalo.edu!callisto.acsu.buffalo.edu!junxu From: Jun Xu Newsgroups: sunyab.cse.740 Subject: Re: MOD-HEAD CASE FRAME Date: Mon, 19 Apr 2004 11:06:36 -0400 Organization: The University at Buffalo Lines: 71 Sender: junxu@buffalo.edu Message-ID: References: NNTP-Posting-Host: callisto.acsu.buffalo.edu Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Trace: prometheus.acsu.buffalo.edu 1082387200 4840 128.205.7.122 (19 Apr 2004 15:06:40 GMT) X-Complaints-To: abuse@buffalo.edu NNTP-Posting-Date: Mon, 19 Apr 2004 15:06:40 +0000 (UTC) X-Sender: junxu@callisto.acsu.buffalo.edu In-Reply-To: Xref: acsu.buffalo.edu sunyab.cse.740:114 Because you didn't define "mod" as a relation. You should add "mod" and "head" relations in "rels" file. On Sat, 17 Apr 2004, Timaporn wrote: > So it means that if I want to represent > "toy gun" = (build head toy mod gun)? > > Another question, I try to put this representation in my demo > but I got the error message "SNePS ERROR: mod is not a relation > Occurred in module relation set evaluatior in function rseval" > What should I do? Do I have to run any commands or represent some relations > first? > > Thanks, > Timaporn > > "William J. Rapaport" wrote in message > news:c5pr1a$ll4$1@prometheus.acsu.buffalo.edu... > > ------------------------------------------------------------------------ > > Subject: MOD-HEAD CASE FRAME > > ------------------------------------------------------------------------ > > > > Several of you have asked for the syntax and semantics of the mod-head > > case frame. > > > > There isn't any official one, and we have been using it in a rather > > informal way. Making it precise would probably be the topic of a > > full-blown master's-level project. > > > > The general idea is to use it to represent compound linguistic > > constructions that are ambiguous in nature. Compare, for a moment, > > the infamous object-rel-possessor case frame that many of you have come > > to know and "love" :-). The idea behind that one is that the > > possessive construction in English (as in: Bill's book, her hat, > > etc.) is a single, compound linguistic construction that is > > variously used to express ownership, part-whole, kinship, and many > > other relations. Since a parser wouldn't necessarily have the > > information necessary to interpret each such occurrence correctly, > > we handle them by representing that single English construction with > > a single case frame, leaving to background knowledge any rules that > > are needed for the full semantic interpretation. > > > > There are other such compound constructions. The most obvious is the > > adjective-noun noun phrase: red hat, small elephant, toy gun. Each > > of these should be represented as a structured individual in SNePS, > > but each has a very different semantics: a red hat is both a hat > > and red; but a small elephant is an elephant, yet not small > > (although it *is* small for an elephant); and a toy gun is a toy, > > but not a gun. Instead of SNePS having to know ahead of time how to > > represent each such expression, we can use the mod-head case frame > > as a "neutral" representation, leaving to background knowledge the > > task of deciding what kind of Adj+N construction it is. > > > > We also tend to use the mod-head construction for other situations; > > perhaps we shouldn't. > > > > So, here's a first attempt at a case frame (I'll ask Stu if he has a > > better idea): > > > > [[(build mod x head y)]] = a structured individual consisting of an > > individual [[y]] modified by [[x]]. > > > > That's pretty vague, but then the mod-head case frame is, too. > > > >