From LOJBAN%CUVMB.bitnet@YaleVM.YCC.YALE.EDU Sat Mar 6 22:55:30 2010 Received: from YALEVM.YCC.YALE.EDU by MINERVA.CIS.YALE.EDU via SMTP; Sat, 27 Mar 1993 03:04:52 -0500 Received: from CUVMB.CC.COLUMBIA.EDU by YaleVM.YCC.Yale.Edu (IBM VM SMTP V2R2) with BSMTP id 9514; Sat, 27 Mar 93 03:03:34 EST Received: from CUVMB.BITNET by CUVMB.CC.COLUMBIA.EDU (Mailer R2.07) with BSMTP id 2690; Sat, 27 Mar 93 01:51:19 EST Date: Sat, 27 Mar 1993 16:48:37 +1000 Reply-To: Nick Nicholas Sender: Lojban list From: Nick Nicholas Subject: Re: TECH: grammar updates X-To: cowan@SNARK.THYRSUS.COM X-Cc: Lojban Mailing List To: Erik Rauch In-Reply-To: from "John Cowan" at Mar 26, 93 11:42:04 am X-From-Space-Date: Sat Mar 27 03:04:52 1993 X-From-Space-Address: @YaleVM.YCC.YALE.EDU:LOJBAN@CUVMB.BITNET Message-ID: The Seraphim surely rejoiced when John Cowan spoke thus: {.i lai seraFIM. cu ja'o salgei ja'ejoica lenu la djan. kau,n. cusku lesedu'u go'e} }la nitcion. poi me la prolog. lojbo cu cusku di'e }> [Dropping "*mo'u"] leaves us with little need for {mau} at all. }> I hope you all are aware of this... (debate welcomed, but i'll have little }> time to participate in it... }Good point. I could live without "mau" and "me'a", and then we'd have two }free CVV monosyllables! Ouch! This one's too drastic, I fear. They're too entrenched in the language. Whjat we can do is make a stylistic recommendation, whereby we point out that a {zmadu} selbri says what we actually usually intend to say by {mau}. }> though I *have* successfully implemented }> tanru-unit-2 -> SE ... BRIVLA, and I allow cmene and brivla to be passed }> onto Prolog from yacc directly, without needing to be predefined within }> Prolog.) }Hurrah, hurrah! I meant, of course, not just parse the beast, but switch the sumti around accordingly. Since the parser will not know all the sumti at the point of parsing tanru_unit_2 (it's yet to parse those following), I had to do some hairy forcing calls to be delayed in the sumti-juggling clauses; cost me a day. But it's done! It just might not run on *your* Prolog compiler! (I'm using 'when' clauses). }> I would have though something like }> tanru_unit_2 -> anyword [(ZEI anyword) ...] ZEI tanru_unit_2 }> to be more sensible: } }That would lead to a mixing of levels. "zei" compounding is done in the }preparser, which doesn't understand nonterminals like "tanru-unit-2"; }"xy. zei se broda" != "xy. zei selbroda" (oddly, both are grammatical; }the first is a tanru with [bogus] seltanru "xy. zei se"). } }OTOH, I could restrict the final term of a "zei" compound to be a brivla, }which I suppose is what you really want. Hm. Given I'd want people to be able to say things like {xy zei go'i}, I suppose that's not warranted: best to let the grammar overproduce than underproduce... }> I find it hard to think of a place structure for }> {xy zei zy}. }So do I. I'm sure someone will work something out :) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Nick S. Nicholas, "Rode like foam on the river of pity Depts. of CompSci & ElecEng, Turned its tide to strength University of Melbourne, Australia. Healed the hole that ripped in living" nsn@{munagin.ee|mundil.cs}.mu.oz.au - Suzanne Vega, Book Of Dreams ______________________________________________________________________________