From @YaleVM.YCC.YALE.EDU:LOJBAN@CUVMB.BITNET Sat Oct 23 05:56:08 1993 Received: from ELI.CS.YALE.EDU by NEBULA.SYSTEMSZ.CS.YALE.EDU via SMTP; Sat, 23 Oct 1993 09:56:55 -0400 Received: from YALEVM.YCC.YALE.EDU by eli.CS.YALE.EDU via SMTP; Sat, 23 Oct 1993 09:56:51 -0400 Message-Id: <199310231356.AA02170@eli.CS.YALE.EDU> Received: from CUVMB.CC.COLUMBIA.EDU by YaleVM.YCC.Yale.Edu (IBM VM SMTP V2R2) with BSMTP id 5191; Sat, 23 Oct 93 09:54:49 EDT Received: from CUVMB.COLUMBIA.EDU by CUVMB.CC.COLUMBIA.EDU (Mailer R2.07) with BSMTP id 3673; Sat, 23 Oct 93 09:57:47 EDT Date: Sat, 23 Oct 1993 09:56:08 -0400 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: Re: TECH: nested bridi anaphora X-To: nsn@MULLIAN.EE.MU.OZ.AU X-Cc: lojban@cuvmb.cc.columbia.edu To: Erik Rauch Status: RO X-Status: This is a known hole in the language, but alas is not so easily solved. I gather from quick reading that you are proposing counting and subscripting like ri, which I think everyone agrees is a pain in the behind and not really practical at normal speech speeds, even if you CAN agree how to count. It is also somewhat redundant to go'i, which is also a ri-like counter, but which skips sub-bridi. It is fairly easy to come up with a sentence that makes ri type counting unbearably hard (not to mention leaving open the question of whether bridi in descriptions shouldn't be counted too - you might have wanted to say the equivalent of "I am the dog, and I can see you"). But better and more common examples are those in which there are multiple levels of abstractions in one sumti (two levels seems to happen a lot), extra bridi stuck in through relative clauses (which may in turn have abstractions), etc. We realized early on that there was no scheme that could handle all the strange cases, and still be usable for the easy cases. Indeed even within sumti we have uncountable values: other sumti within the current sumti of the main bridi are not counted in "ri" reference. It is easy to come up with situations where they are useful - take any bridi with a ri expressed, and put it in an abstraction sumti, and the "ri" refers to the first sumti before the abstraction sumti and not to the other place within the abstraction (i.e. la djan. nelci lenu le mlatu cu tancysatre ri says that John likes the cat to lick _him_, as opposed to the cat itself) That is why the little-used lerfu anaphora may be a better long-term solution, especially for complex cases. There will be times when ri/ra/ru will be very clear, other times when vo'a/vo'e will work, and ke'a seems to work very nicely in its limited domain. But lerfu are better than ko'a in that they needn't be assigned to refer, provided that there is only one plausible sumti starting with the appropriate letter. (The tradeoff is that they have lower redundancy - ty and py could be easily confused. But I guess you could use "anyword bu" to solve that one.) Turning back to the bridi problem, I never saw go'i and family as being even as powerful as the set of sumti anaphora. I put in what became no'a patterned after ke'a, for use in relative clauses. I can't recall when "nei" came in, but it is recent (since 6/90, when "nei" was freed as a selma'o related to lerfu) and the definition right now is murky - is it the main bridi of the current sentence (as I thought, since I vaguely recall a cmavo patterning after vo'a-series) or the sub-bridi that you are working in right now. John can probably clarify when it is supposed to be used in his concept (it doesn't matter too much - one is no'ixino and the other no'ixiro, but we oughta figure out which is most useful or get rid of it). But given that go'i itself is not powerful enough to handle all cases, I am inclined to go to our systems of anaphora that are more like ko'a and lerfu. The equivalent of ko'a are the brodX series, which presumably should be assigned with "cei". We have no obvious equivalent to lerfu, though it seems that you could use brodaxiBY to get the point across. But all of this seems a waste. Why, in the examples you present, not just use ellipsized viska and zerle'a, both of which are the same or fewer syllables as the cmavo solutions you propose? I'm having trouble seeing a pattern of situations that a new cmavo would handle where the result would be consistently shorter than the typical repeating of the referent. Maybe when the language has a lot of 8 syllable brivla this will seem more useful, but I suspect that such restricted brivla might often be anaphorized by using the gismu in the final position as an abbreviation, and you always have subscripted broda to fall back on. (A series of nonsense gismu starting with each consonant sounds interesting, too, as an analogue to the lerfu anaphora - but I don't see proposing it unless we find usage demanding it.) I see this "hole" in the language as being of a different order than the "missing demonstratives", which can only be approximated by other wordings. lojbab