From LOJBAN%CUVMB.BITNET@UBVM.CC.BUFFALO.EDU Sat Nov 4 22:01:08 1995 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 WAA01906 for ; Sat, 4 Nov 1995 22:01:05 -0500 Message-Id: <199511050301.WAA01906@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 4BDB1129 ; Sat, 4 Nov 1995 22:56:53 -0400 Date: Sat, 4 Nov 1995 18:46:01 -0800 Reply-To: "John E. Clifford" Sender: Lojban list From: "John E. Clifford" Subject: rafsi To: lojban list Status: OR Well, at least mark,l is willing to do some of the work his broad complaint suggested. As he has pointed out, the system he is working with is a rafsi and lujvo system, so we must withhold final judgement until we see in detail how some gaps are filled in and how the glue rules are worked out in detail. We can make a few preliminary remarks however, perhaps constructive in suggesting how the those rules ought to be formulated. First of all, the plan does get rid of the rafsi that look like cmavo, though this was at most an irritant for people who learn vocabulary from a list rather than in context. In particular, it was never an ambiguity (and not strictly even a case of homophony). But it was a flaw in many an aesthetic view of systems, so worth fixing at a small cost. Secondly, it does give every (or pretty near every -- there are some unproven cases) a short rafsi. Many of these are, of course, four letters long though still one syllable. So, at least in theory, most two-gismu lujvo could be two syllables. In fact, since the new rafsi create new contact cases, most of these will actually be three or even four syllables when the work is done. The average lujvo will probably get slightly longer on this proposal (almost certainly on letter count, probably also on syllable count). But it will still be (slightly again) shorter than a system that used only the current four- and five-letter rafsi and the current glue. (The system also provides for over half the current gismu another of mark,l's desiderata, a relatively easy and uniform way to get from gismu to rafsi and back, like cases being treated alike and all. A pure bonus, of course, and not likely to carry over to the remaining cases, the assignment of CAC and STA and STEI and SPIA forms.) The space this proposal makes available for word final affixes is probably adequate, even without a lot of tarting up. The present system uses only about 625 word- final rafsi (CVV,CV'V, CCV) and that already includes several cmavo imported into the rafsi system. It does, of course, have more room for expansion that the proposed system, even with the tarting up. But, assuming that the space can be fully utilized, it should be as effective as the present system in its present form. The assumption there is a problem. I suppose mark,l wants to assign those spaces to the gismu most likely to end up as the base of compounds. But, as he has pointed out, figuring out which those are is not an easy task, though he will have a lot more Lo??an data than either JCB or Lojbab had. It will still be unsatisfying to many people all the time and to each person at least some of the time. So he will do the best he can according to his sense of rightness, as everyone else has done before him, and fare about as well as they -- or better, given the extra data (though still largely American English at its heart). The price to pay for an occasional misuse of an affix here is minor. The problem with the CAC forms is more severe, since here it is possible he may lose an gismu altogether -- maybe not an important one, but enough to drop the claim of complete coverage (not to mention the various symmetries from earlier desiderata, but they were doomed anyway). For presumably even these forms are stuck with the Cs and Vs of the original and it is at least theoretically possible that there could be a set of gismu such that no combination of forms could cover them all. There probably is not such a set in the present list and care could be taken in expanding to not allow one (indeed, such care may have been taken even in the present list) but, until the task of assigning is done, it remains a possibility. A more likely possibility is that some legal CAC has no legal gismu to go to -- or none that is not more obviously related to another CVC. So the full space may not get used, throwing the statistics off a bit. But also leaving room for some expansion -- not a bad thing. This latter might also happen with the word-final space, the flip side of moving some things from four-letter space to three letter space. The proposed system would need (assuming that the sticking together of rafsis is not done in any surprising way) all of the current rules about hyphens of various sorts plus some others, as mark,l notes, for new contacts, almost all of which would apparently need y's of the like, for pronunciation, if not for resolution. I think it unlikely that the resolution algorithm would require more glue with these forms -- when they differ from long-affix forms in the present system, they seem to be on the side of even more safely brivla forms. The le'avla/fu'ivla space would also need also need redefinition, though the old vague "anything that can pass as a brivla but is not a gismu or lujvo" would still work -- but would be a smaller class, lujvo space having expanded. Whether the internal resolution of a lujvo into a unique string of rafsi would always be be unique needs someone with a quick hand at combinatorics to work through. With a larger number of rafsi types, the proof will be more complex to complete unless there is a quicky guarantee of success in the structure somewhere -- as the various structural descriptions suggest there may be. The last question before we might get around to considering this proposal is, is it worth the effort? After seeing the advantages or breakeven points above, consider the following objections. The problem it officially sets out to solve is of no particular significance to the language, however annoying it is to a learner. The solution ends up making lujvo slightly longer, even than they would be with the perfect coverage that the present system allows. It gives prime rafsi real estate to rarely used forms (the cmavo), an aesthetic minus in many people's eyes -- certainly worse that the original problem of rafsi/cmavo doublets. I suspect each of us could add a few favorites of our own. But then each must do the final toting up. But do wait for the final form of the proposal, with all the proofs and rules in place (though the need for new rules and proofs may turn a few off already). pc>|83