Path: utzoo!utgpu!attcan!uunet!husc6!mailrus!purdue!decwrl!ucbvax!CS.ROCHESTER.EDU!nl-kr-request From: nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) Newsgroups: comp.ai.nlang-know-rep Subject: NL-KR Digest Volume 5 No. 23 Message-ID: <8811030524.AA23950@teak.cs.rochester.edu> Date: 3 Nov 88 05:10:00 GMT Sender: daemon@ucbvax.BERKELEY.EDU Reply-To: nl-kr@CS.ROCHESTER.EDU Organization: University of Rochester, Department of Computer Science Lines: 456 Approved: nl-kr@cs.rochester.edu NL-KR Digest (11/03/88 00:08:53) Volume 5 Number 23 Today's Topics: Re: Syntactical *definition* of English talking to Unix (Re: Syntactical *definition* of English) Submissions: NL-KR@CS.ROCHESTER.EDU Requests, policy: NL-KR-REQUEST@CS.ROCHESTER.EDU ---------------------------------------------------------------------- Date: Wed, 19 Oct 88 12:10 EDT From: Tim Budd Subject: Re: Syntactical *definition* of English You may remember that Context Free Languages were discovered by a Linguist, Noam Chomsky, not a computer scientist. At the time (mid 1950's), there was great hope that a CFL, or at worst a CSL (context sensitive language) could be found that would describe English, and other such grammars developed for other natural languages. Such efforts more or less met with utter and complete defeat in the late 50's and 60's. Indeed so much so that some people working in understanding English (such at the folks at Yale), almost totally abandoned any notion of syntax, and proceeded with just a semantic analysis of utterances. So I fear your quest will be a futile one; the best you can hope for is a grammar for a rather stilted and minimal subset of English. ::= ::= I | teachers | policemen | the mob ::= eat | love | detest ::= mice | chocolate | teachers | little children ------------------------------ Date: Wed, 19 Oct 88 16:15 EDT From: Ralph Hyre Subject: Re: Syntactical *definition* of English In article <6946@orstcs.CS.ORST.EDU> budd@mist.UUCP (Tim Budd) writes: >Linguist, Noam Chomsky, not a computer scientist. (context sensitive language) could be found that would describe English, and >other such grammars developed for other natural languages. even stilted English would be enough for me. I just want to talk to my Unix system in a more converstational manner, I have having the keystrokes 'ls -al' burned into my brain, wasting those valuable neural pathways. >Such efforts more or less met with utter and complete defeat in the late >50's and 60's. Interesting that some of the technology lived on in the educational system: (ie my school system) 'phonics' (the name given to my 3rd grade language class), where we learned S -> N V, and more elaborate sentence diagramming in 7th grade: S -> NP VP, NP -> prep N, N -> cat,dog, prep -> about, above & 50 others. Then, in college, I learned about REAL linguistics and affix hopping and such. -- - Ralph W. Hyre, Jr. Internet: ralphw@ius3.cs.cmu.edu Phone:(412) CMU-BUGS Amateur Packet Radio: N3FGW@W2XO, or c/o W3VC, CMU Radio Club, Pittsburgh, PA "You can do what you want with my computer, but leave me alone!8-)" ------------------------------ Date: Wed, 19 Oct 88 23:34 EDT From: Greg Lee Subject: Re: Syntactical *definition* of English From article <6946@orstcs.CS.ORST.EDU>, by budd@mist.cs.orst.edu (Tim Budd): " You may remember that Context Free Languages were discovered by a " Linguist, Noam Chomsky, not a computer scientist. At the time (mid " 1950's), there was great hope that a CFL, or at worst a CSL (context " sensitive language) could be found that would describe English, and " other such grammars developed for other natural languages. " Such efforts more or less met with utter and complete defeat in the late " 50's and 60's. Indeed so much so that some people working in understanding Context free phrase structure grammar lives! It's the basis of the best current theory of syntax, GPSG -- Generalized Phrase Structure Grammar. Greg, lee@uhccux.uhcc.hawaii.edu ------------------------------ Date: Thu, 20 Oct 88 00:55 EDT From: Rob Bernardo Subject: Re: Syntactical *definition* of English In article <3349@pt.cs.cmu.edu> ralphw@ius3.ius.cs.cmu.edu (Ralph Hyre) writes: +even stilted English would be enough for me. I just want to talk to my +Unix system in a more converstational manner, I have having the keystrokes +'ls -al' burned into my brain, wasting those valuable neural pathways. Let's see, if your UNIX system understood conversational English only, you'd have to say: Give me a long listing of everything in the directory. -- Rob Bernardo, Pacific Bell UNIX/C Reusable Code Library Email: ...![backbone]!pacbell!rob OR rob@PacBell.COM Office: (415) 823-2417 Room 4E750A, San Ramon Valley Administrative Center Residence: (415) 827-4301 R Bar JB, Concord, California ------------------------------ Date: Thu, 20 Oct 88 06:09 EDT From: Clay M Bond Subject: Re: Syntactical *definition* of English Tim Budd: >You may remember that Context Free Languages were discovered by a >Linguist, Noam Chomsky, not a computer scientist. No, I don't, actually. I'm not quite sure what you mean here. They certainly weren't "discovered" though if this is supposed to mean that Nim first proposed that natural language could be generated with a CFG then it makes more sense (though that, too is wrong. Harris, not Nim.) >1950's), there was great hope that a CFL, or at worst a CSL (context >sensitive language) could be found that would describe English, and You mean CF/SG, don't you? If language X can be generated by a CFG, then language X is a CFL; a CFL is not going to describe English. >Such efforts more or less met with utter and complete defeat in the late >50's and 60's. No argument. >Indeed so much so that some people working in understanding >English (such at the folks at Yale), almost totally abandoned any >notion of syntax, and proceeded with just a semantic analysis of >utterances. I fail to see what the difference is, assuming the semantic analyses used are mathematical possible-worlds models which have nothing to do with reality, much less language. You're manipulating symbols. How is manipulating semantic symbols different from manipulating syntactic ones, save that the former is more challenging since it's more obvious that symbol systems don't work. What? This construction doesn't fit the rule? Write another rule/feature, of course! The plight of the semanticist is no less futile than the syntactician. -- <<<<<<<<<<<<***<<<<<<<<<<<<***<<<<<<***>>>>>>***>>>>>>>>>>>>***>>>>>>>>>>>> << Clay Bond, IU Department of Leather er uh, Linguistics >> << ARPA: bondc@iuvax.cs.indiana.edu AKA: Le Nouveau Marquis de Sade >> <<<<<<<<<<<<***<<<<<<<<<<<<***<<<<<<***>>>>>>***>>>>>>>>>>>>***>>>>>>>>>>>> ------------------------------ Date: Thu, 20 Oct 88 13:56 EDT From: Kevin S. Van Horn Subject: Re: Syntactical *definition* of English In article <6946@orstcs.CS.ORST.EDU> budd@mist.UUCP (Tim Budd) writes: >Such efforts more or less met with utter and complete defeat in the late >50's and 60's. Indeed so much so that some people working in understanding >English (such at the folks at Yale), almost totally abandoned any >notion of syntax, and proceeded with just a semantic analysis of >utterances. So I fear your quest will be a futile one; the best you >can hope for is a grammar for a rather stilted and minimal subset of >English. I think that Fred Thompson, of the Caltech C.S. Dep't., would not entirely agree with this statement. His work is in natural-language interfaces and, though recognizing its limits, he has managed to do quite a bit using a syntax-based approach. The person who originally asked about this may want to write Dr. Thompson, at Caltech 256-80, Pasadena, CA 91125. Kevin S. Van Horn ------------------------------ Date: Thu, 20 Oct 88 16:24 EDT From: Dave Lawrence Subject: Re: Syntactical *definition* of English rob@pbhyf.PacBell.COM (Rob Bernardo) writes: >ralphw@ius3.ius.cs.cmu.edu (Ralph Hyre) writes: >+even stilted English would be enough for me. I just want to talk to my >+Unix system in a more converstational manner, I have having the keystrokes >+'ls -al' burned into my brain, wasting those valuable neural pathways. > >Let's see, if your UNIX system understood conversational English only, >you'd have to say: > > Give me a long listing of everything in the directory. or, more accurately, you would have to tell it Give me a long listing (permissions, groups and all that good stuff) of every file in the -current- directory. (unless you had a parser that understood implied words ...) Wouldn't you just love to write the parser that could correctly handle, in the English (not -American- (personal pet peeve) |:-) language the equivalent of the following... alias news-dates grep 'Date:' /usenet/spool/\$1/* | sed 's/.*:.*: \(.*\)/\1/' | sed 's/^. / &/' | sort | sort -f -M +1 | sed 's/\(.* \)..:.*$/\1/' | uniq -c Well, it might not look quite as bad, but I wouldn't say it to mum at Christmas dinner .... Cheerio, Dave -- g l o r i o u sex i s t e n c e EMAIL: tale@rpitsmts.bitnet, tale%mts.rpi.edu@rpitsgw, tale@pawl.rpi.edu ------------------------------ Date: Thu, 20 Oct 88 19:27 EDT From: Steven Ryan Subject: Re: Syntactical *definition* of English >You may remember that Context Free Languages were discovered by a >Linguist, Noam Chomsky, not a computer scientist. At the time (mid >...... Eh? I think somebody forgot Type 0 = Turing Machine. Anyway, check out Appendix ?B of Terry Winograd's book, some or other, Part I: Syntax. No, nobody has a complete, formal syntax/semantics of any natural language, but, you said you wanted it for a game? this kind of stuff covers most cases. For what it doesn't, just respond Eh? I'm sorry, I don't understand; could you repeat that using simpler sentence? ------------------------------ Date: Fri, 21 Oct 88 15:20 EDT From: Rick Wojcik Subject: Re: Syntactical *definition* of English In article <2509@uhccux.uhcc.hawaii.edu> lee@uhccux.uhcc.hawaii.edu (Greg Lee) writes: >Context free phrase structure grammar lives! It's the basis of the >best current theory of syntax, GPSG -- Generalized Phrase Structure >Grammar. > Greg, lee@uhccux.uhcc.hawaii.edu Greg, I would be interested in knowing the criteria by which you judge one 'current theory' of syntax to be better than the others. Why is GPSG better than HPSG, in your opinion? Than LFG? (Don't bother with GB. I don't want to stir up trouble. :-) -- Rick Wojcik csnet: rwojcik@boeing.com uucp: uw-beaver!ssc-vax!bcsaic!rwojcik ------------------------------ Date: Sat, 22 Oct 88 08:12 EDT From: Jay Kim Subject: Re: Syntactical *definition* of English > <<<<<<<<<<<<***<<<<<<<<<<<<***<<<<<<***>>>>>>***>>>>>>>>>>>>***>>>>>>>>>>>> Clay Bond wrote: > a CFL is not going to describe English. Could you tell us a convincing evidence for this? If you are going to bring up 50's argument based on a long-distant dependency, I would recommend you to read first Gerald Gazdar (1982) Phrase structure grammar. In Pauline Jacobson and Geoffrey K. Pullum (eds), The Nature of Syntactic Representation. Dordrecht: D. Reidel, 131-186. ------------------------------ Date: Sat, 22 Oct 88 21:54 EDT From: Ralph Hyre Subject: talking to Unix (Re: Syntactical *definition* of English) I've been flamed by about 6 different people for wanting to be more conversational when talking to my Unix machine. I understand the point about verbosity, and I wouldn't ALWAYS exclusively deal with it on that basis. The NeXt machine will probably be able to support limited vocabulary speech recognition in real time. Maybe, eventually, well enough so that I can say 'find the disk hogs and flame them', and the user interface will be able to figure out what commands need to be run. If not, it can ask me. At that point, I don't really care what OS is sitting underneath, I just picked Unix as a well-known example. I want maximal effective synergy between the input modalities I use [sounds, images, and muscles (typing).] >Wouldn't you just love to write the parser that could correctly handle, >in the English (not -American- (personal pet peeve) |:-) language the >equivalent of the following... >alias news-dates grep 'Date:' /usenet/spool/\$1/* | sed 's/.*:.*: \(.*\)/\1/' | sed 's/^. / &/' | sort | sort -f -M +1 | sed 's/\(.* \)..:.*$/\1/' | uniq -c I don't care who or what writes it, I just want to be able to USE it someday. Anyway, I just wanted to clarify that. -- - Ralph W. Hyre, Jr. Internet: ralphw@ius3.cs.cmu.edu Phone:(412) CMU-BUGS Amateur Packet Radio: N3FGW@W2XO, or c/o W3VC, CMU Radio Club, Pittsburgh, PA "You can do what you want with my computer, but leave me alone!8-)" ------------------------------ Date: Sun, 23 Oct 88 06:26 EDT From: Greg Lee Subject: Re: Syntactical *definition* of English From article <8330@bcsaic.UUCP>, by rwojcik@bcsaic.UUCP (Rick Wojcik): " Greg, I would be interested in knowing the criteria by which you judge one " 'current theory' of syntax to be better than the others. Why is GPSG " better than HPSG, in your opinion? Than LFG? (Don't bother with GB. I " don't want to stir up trouble. :-) Actually, it's only context free phrase structure grammar I'm prepared to defend, not GPSG specifically. The nice thing about GPSG is that a GPSG description abbreviates a finite number of CF phrase structure rules, and so describes a context free language. If and to the extent the other theories you mentioned allow a similar interpretation, I love them, too. But I don't know whether they do. I should admit that I find much of the current literature in syntax difficult to understand, since though it purports to be about syntactic theory, it seems really only to concern conciseness or convenience of description. This includes GPSG, the book, by Gazdar, Klein, Pullum, and Sag. To what I said in reply to Walter Rolandi, I'd like to add something about the local nature of lexical subcategorization, again, following Gazdar. Subcategorization of items with respect to sister constituents is straightforward in a context free phrase structure grammar, and this is the only, or at least the predominate, kind of subcategorization found in natural language. However, I'm not sure it's possible to make this out as a prediction of CFPSG without an appeal to simplicity, since one can also describe certain non-local subcategorizations. Greg, lee@uhccux.uhcc.hawaii.edu ------------------------------ Date: Mon, 24 Oct 88 18:50 EDT From: Don Steiny Subject: Re: Re: Syntactical *definition* of English / hpcupt1:sci.lang / budd@mist.cs.orst.edu (Tim Budd) / 9:10 am Oct 19, 1988 / Such efforts more or less met with utter and complete defeat in the late 50's and 60's. Hmm, read: "Lectures on Government" and Binding by Noam Chomsky. TG lives!! ------------------------------ Date: Mon, 31 Oct 88 15:46 EST From: Rick Wojcik Subject: Re: Syntactical *definition* of English Greg Lee has praised GPSG on the grounds that it distinguishes "a category difference between whole constituents and constituents from which something has been extracted..." In response to my claim that we have little or no use for 'relativized relative clauses', he points to the phenomenon of resumptive pronouns: GL> I think you're going to encounter some difficulty with GL> dialects/languages having resumptive pronouns: 'The man who I met GL> the girl that knew him ...'" I'm glad that you raised this point, Greg. Can you explain to me how GPSG is able to handle the phenomenon of resumptive pronouns in such a way that their properties are naturally distinct from gaps? Note that relative clauses that violate extraction constraints *must* contain pronouns that agree with the head NP. How is this done? Through feature percolation? So why don't gaps percolate in the same way? It seems to me that resumptive pronouns override extraction constraints because they are more salient than gaps. The more reasonable treatment of the resumption/gap differences seems to lie in the direction of behavioral properties, not categorical distinctions between gapped and ungapped constituents. -- Rick Wojcik csnet: rwojcik@boeing.com uucp: uw-beaver!ssc-vax!bcsaic!rwojcik ------------------------------ Date: Wed, 2 Nov 88 08:30 EST From: Greg Lee Subject: Re: Syntactical *definition* of English From article <8457@bcsaic.UUCP>, by rwojcik@bcsaic.UUCP (Rick Wojcik): " ... " I'm glad that you raised this point, Greg. Can you explain to me how GPSG is " able to handle the phenomenon of resumptive pronouns in such a way that their " properties are naturally distinct from gaps? I doubt that I can. I don't know much about resumptive pronouns. " Note that relative clauses that " violate extraction constraints *must* contain pronouns that agree with the " head NP. Two counter-notes. Such relative clauses do not violate extraction constraints, strictly speaking, since nothing is extracted (obviously). And it is not clear that they must contain the pronouns you say they must. After all, there are relative clauses that contain neither gaps nor resumptive pronouns -- we call them appositive. " How is this done? Through feature percolation? So why don't gaps " percolate in the same way? If the occurrence of relative clauses with resumptive pronouns does have to be syntactically constrained in this sort of way (not clear), then that's a real problem. As you suggest, I would be confronted with finding a principled way to distinguish the two kinds of features or two kinds of percolation. I think my prospects would be dim. " It seems to me that resumptive pronouns override " extraction constraints because they are more salient than gaps. The more " reasonable treatment of the resumption/gap differences seems to lie in the " direction of behavioral properties, not categorical distinctions between " gapped and ungapped constituents. You're making distinctions that I don't know how to make. It's behavioral properties under discussion in any case, I would have thought. Greg, lee@uhccux.uhcc.hawaii.edu ------------------------------ End of NL-KR Digest *******************