From - Mon Mar 04 09:38:59 1996 Received: from VMS.DC.LSOFT.COM (vms.dc.lsoft.com [205.186.43.2]) by locke.ccil.org (8.6.9/8.6.10) with ESMTP id LAA17223 for ; Sun, 3 Mar 1996 11:42:12 -0500 Message-Id: <199603031642.LAA17223@locke.ccil.org> Received: from PEACH.EASE.LSOFT.COM (205.186.43.4) by VMS.DC.LSOFT.COM (LSMTP for OpenVMS v1.0a) with SMTP id 8C5E2639 ; Sun, 3 Mar 1996 11:03:42 -0500 Date: Sun, 3 Mar 1996 11:01:56 -0500 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: Re: TECH: fuzzy: vs. X-To: sbelknap@uic.edu X-cc: lojban@cuvmb.cc.columbia.edu To: John Cowan X-Mozilla-Status: 0011 Content-Length: 8260 >I am trying to piece together how >the selmaho work from the terse descriptions in the cmavo list The terse descriptions are not really meant to be tutorial, but rather as reminders assuming that you have learned the words and selma'o elsewhere. Once youhave learned how a given selma'o works grammatically, then the cmavo definition is usually lenty clear at showing you the semantics of that particular cmavo within the selma'o. >nd from >trying to puzzle out the BNF grammer Something I have never been able to do either. I use the YACC grammar. It is harder to understand the "big picture" with the YACC grammar, but you build pieces that are the equivalent of constructs you already undersytand like, "sumti", "selbri", "sentence", and then try to understand them as wholes of that type. You have the additional advantage of using JCB's old books if you choose, provided you can related the selma'o to JCB's selma'o. In the case of MOI, JCB's equivalent is "-ra" and "-ri" suffixes. > Which of the refgrammer >papers explains how the selmaho classes relate to the YACC grammer or the >BNF? Hunh? THat is the WHOLE refgrammar. Not just one piece, but every single paper. That is in essence what the refgrammar is for, specifically to relate the selma'o to the grammar, and to incidentally provide a little semantics. > There seems to be a set of knowledge out there that people like jorge >have somehow mastered that I can't find. Jorge first mastered the basics of the language by writing a lot of simple stuff. He read and reviewed the refgrammar over a year and a half ago, and has been writing more stuff, and more complex stuff, since then. His knowledge didn;t spring magically from reading a document , but from using a language. Nick mastered the language without any refgrammar, just from the diagrammed summary and the textbook (which covers only part of the grammar, though in greater detail). But he did even MORE writing in the language, and revised everything as he learned more. The people who use the language learn the language, and that is why I think they alone are qualified to say where the language should go post-baseline. Jorge has an immense PRACTICAL knowledge of the language through his heavy use, and when you have that, the selma'o just fit naturally into place. >The predicate so reusulting could be used in tanru, in sei/se'u phrases, >>etc. There is a predicate equivalent to negation/affirmation: jetnu/jitfa >>so you can use a predicate with jetnu to get the equivalent of ja'a with a >>subscript. There are all manner of other areas, some unexplored, where >>predicates can pop into unusual grammatical locations. >> > >Are the precedence orders worked out for all the metalinguistic operators? >I suppose that they are implicit in the BNF and YACC. But it appears that >sei...sehu phrases, for example, would not have the right scope to be very >general. sei/se'u, like all "free modifiers" have a specifically defined scope. You can put the modifer in various places to vary the scope. z>Also, it seems like an ugly kluge to resort to metalinguistic >commentary to implement fuzziness, which should be about as easy to use as >discrete description, in my opinion. Use of loaded vocabulary: "kluge", "resort", suggest tyhat you don't understand that this is a major and basic mechanism in Lojban in order to carefully control scope. The free modifiers were added based on an idea by Carter in order to allowsuch things at various levels. The reason for labelling what you are describing for fuzziness as "metalinguistic" is that all of the examples that you have presented have analyzed down to metalinguistic statements about an otherwise simple predicate. The core of the language is a simple predicate relationship. All of the bells and whistles are either metalinguistic expansion or modification of the simple predicate, or abbreviations of such expansions. Thuse for example, almost every relative clause can be expressed in a separate sentence metalinguistically linked to the original sentence. Likewise every tense and negation. Why should fuzziness be treated DIFFERENTLY is the question. And the other issue is how much abbreviation is to be designed in. If you keep requirements simple, we can use very succinct abbreviation. But what you have generally presented for fuzziness is such a broad and unrestricted concept that there IS no single abbreviated form possible. For example, as I understand what you have written, a measure of fuzziness CAN involve 1) a number indicating a position on a scale, or memebership in one of a number of discrete categories 2) a number indicating the total number of categories, or the range of the scale 3) an indication whether the categories/scale is ordered 4) an indication of what is being scaled (truth, baldness, red vs. green-ness). You could potentially need to express all of these things, which grammatically constitutes two MEX expressions, a connective, and a ni-abstract bridi perhaps with lambdas. And each of those constructs in Lojban is potentially a hairy mess in itself. And you want to be able to glue this mess to pretty near any grammatical unit in a sentence. The ONLY construct that can glue to pretty near any construct in a sentence is a free modifier, or an attitudinal, and attitudinals cannot have subgrammar of the sort that I mentioned above. The ONLY way you will get something brief will be to give up flexibility. You cannot express every possible tense construct, but must expand into multiple sentences with metalinguistic links between them, using the brief tense forms. If you want comprehensive coverage of a realm of logic, you will have complex constructs and few short forms - the only brief stuff in logical expression in Lojban is stuff that has been honed for 40 years since JCB started, and we aren't going to be able to do more than a "kluge" in abbroievauted forms in a few months or even years at devising corresponding short forms for the less well-defined requirements of fuzziness. To do the best we can, thereofre, we need to understand what the requirements are, not in language feature terms, but traced back tot the literature on fuzzyiness that all of us can read, and that some like pc, already understand. There are many things called "fuzzy logic", and pc has said that he thus far cannot clearly identified what you have talked about as being exactly identical with any of the levels and categories of fuzzy logic that he knows, but rather that you are overlapping definitions in way thatmake it hard to see what you are trying for. If you define your requirements ONLY in terms of "how can I say this", then you will get constructs that allow you to say "this", but which may not exapnd to cover what is to you the similar problem "that", because we will not have seen that "this" and "that" are related problems, since you haven't described the problems generally enough. But beware of being TOO general, because as I said above, what the most general interpretation of your requirements now seems to need is a free modifier that can have an extremely rich grammar, but which has no need for any specific-to-fuzziness grammar because extreme complexity in Lojban can ONLY be handled in one way. >How does all this impact vs. ? And seems to be saying what we have said before, that as your reqirements become clearer (and more broad), fuzziness just becomes another kind of predication, and is expressed using a predicate attached in some subordinate way to the main bridi. Only free modifiers, and in particular the SEI/SEhU metalinguistic or TO/TOI parenthetical are of the order necessary to handle everything you seem to be asking to do. SO far, the ONLY thing I am willing to add is Cowan's MOI for fuzzy scales, and as I said that is only because I so long misworded cu'o that people no longer recognize that it was intended to be just that sort of fuzzy-MOI. Otherwise, existing constraycts and cmavo can handle everything you have asked for in sufficiently succinct forms. lojbab