From lojbab Fri May 17 01:53:32 1991 Return-Path: Message-Id: Date: Fri, 17 May 91 01:52 EDT From: lojbab (Bob LeChevalier) To: lojban-list Subject: goi - response to jimc Status: RO jimc writes: >I see here "le nanla goi ko'a" where the main phrase is the antecedent of >the anaphor, and also "ko'e goi lo velcnemu..." where the anaphor is >the main phrase. Is this really allowed? For humans speaking normally >it is fairly easy to distinguish the anaphor from its newly assigned >antecedent, but to nail down the difference in all possible circumstances >may be more difficult. I would recommend that "se" (explicit conversion) >be required on one ordering or the other. >If "goi" may be used to "also-known-as" a pair of non-anaphor sumti, >you have two interpretations of "A goi B": one analogous to the anaphor >case in which one sumti is primary and the other is a symbol or >abbreviation for it, and another where A and B are alternative >specifications of the same referent set each of which can stand on its >own. That's commutative whereas the first interpretation is not. Historical note: jimc's writings on Loglan were the source of "goi", the cmavo for relative-phrase anaphora definition. Yes, both "referent-sumti goi anaphora", and "anaphora goi referent-sumti" are both allowed. I guess theoretically you could have anaphora on both sides, but this would occur only in the pragmatically useless "referent-sumti goi " and its mirror image. To equate alternate specifications for the same referent set, where anaphoric assignment is not being performed, you would use the restrictive appositive "po'u" or the non-restrictive appositive "no'u". An anaphora resolver would not have to be too smart, given the assumption that one of the set of 'equated' sumti is a referent-sumti, and the other(s) are assignable anaphora. First let's list the types of sumti, emphasizing the anaphora: assignable variables in ko'a/fo'a lerfu 'new' le descriptions that have not been previously assigned through usage or speaker/listener familarity with the context semi-assignable mi/do (usually assigned using vocatives) (rarely) la names that have not been previously assigned _______________ non-assignable da-series, elliptical anaphora, ri-series back-counting anaphora, demonstratives, metalinguistic sentence anaphora, etc. in KOhA all other sumti Assuming that one sumti must be the referent, that sumti is the lowest on this hierarchical list: "ko'a goi le sumti" assigns "ko'a", etc. The only tricky thing for a computer will be contextual "le" assignments, and these are risky and liable for confusion if unmarked anyway ("ca'e" - lexeme UI - could perhaps be used to resolve an ambiguous assignment situation). An example of such a case might be "lemi bersa goi le nanla" = "My son, hereinafter 'the boy'". Of course the ultimate resolution in such ambiguous situations is to see which of the two IS used later as an anaphora - they should not both be used, since this would violate the semantic sense of anaphora assignment. In any case, we're pushing what will plausibly be done with goi anyway - I've never used it except to assign ko'a/fo'a. Anaphora may be reassigned of course, and things could get trickier, but I think the above hierarchy indicates when this is easily possible. There is a member of lexeme UI that resets all assigned anaphora to 'undefined'. "ni'o" releases all assignments; however, a footnote - If you started the text with a longer-than-one string of "ni'o"s, anaphora assigned at that hierarchical scope are not freed by a shorter string of "ni'o"s. This allows long texts to be broken up into an outline. However, I suspect under pragmatic usage, each chapter would reset all anaphora and then immediately repeat/summarize a list of 'continuing anaphora' and their assignments. This is only fair to the reader - even in English prose where 'he' can only refer to 'Bob', the name is occasionally repeated to remind the listener who 'he' is. These summarized anaphora reassignments will perhaps be a challenge to a computer because they will no doubt occasionally use elliptical references to the context that require real-world knowledge. e.g. john is the father of sam. ko'a goi le bersa ... (x1 - the son ...) requires that le bersa be interpreted even though never 'defined' as an anaphora for 'sam'/'le se patfu'. (There are also bridi anaphora - lexeme BUhA, corresponding to the existential da-series, pseudo-lexeme BRODA in the gismu space, corresponding to the assignable free-variable ko'a-series, and GOhI, which has a variety of others. "cei"/Lexeme CEI is the equivalent of "goi" for assigning bridi anaphora. The only defined usage would have a BRODA member on one side, although there have been some discussions of using CEI in prenex restrictions of BUhA.) lojbab