From LOJBAN%CUVMB.BITNET@UBVM.CC.BUFFALO.EDU Tue Nov 7 13:32:40 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 NAA11554 for ; Tue, 7 Nov 1995 13:32:33 -0500 Message-Id: <199511071832.NAA11554@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 EFA63D87 ; Tue, 7 Nov 1995 14:29:40 -0400 Date: Tue, 7 Nov 1995 10:15:38 -0800 Reply-To: "John E. Clifford" Sender: Lojban list From: "John E. Clifford" Subject: rafsi To: lojban list Status: OR mark,l Homophone affix ambiguity -- the fact that hundreds of short rafsi are identical to cmavo with unrelated meanings -- does not lead to the creation of ambiguous sentences. The ambiguity does not "reach" to the sentence "level." You can always disambiguate before the sentence is finished. pc: This, of course, is a misleading way of saying "There is no ambiguity here", since an ambiguity only exists if it makes a difference in utterance meaning. What is left is a minor processing slowdown -- not even so long a one as deciphering a compound *known to be a compound*: mark,l: You're listening to a Lojban utterance (or reading a Lojban text). Along the stream of speech (or string of text) comes a form (such as CVV or CV'V) that you recognize immediately as being meaningful in Lojban. Yet you cannot know its meaning as quickly, if it's one of the 295 such forms with two unrelated meanings; you must hesitate between the two possible meanings. Meanwhile, the stream (or string) keeps flowing. As soon as you hit a word boundary -- judging this by stress in speech (or by spacing in text) -- you can disambiguate. If the form stood alone it was a cmavo, so you settle on the cmavo meaning. If the form was part of a compound, you must reckon whether the compound was a lujvo or a compound cmavo -- judging this by the presence or absence of a consonant pair (or of -r- or -y- hyphenation). If the compound was a lujvo, then you settle on the rafsi meaning. pc: Recheck the resolution algorithms again. You usually know that something is a rafsi immediately and the doubt never lasts longer than two phones -- less than one morph, let alone to word boundary. If you are in the midst of a brivla, between (inclusive) a CC and primary stress, then any CVV is a rafsi. Otherwise it is one exactly if it is followed imediately by rC or nr (at the beginning of a lujvo). All remaining cases are cmavo. This does not represent a perceptable delay in the interpretation process. Learning rafsi as part of learning gismu and learning cmavo separately -- or learning both directly from running text -- remove the purported learning problems with the "homophones;" those problems look like the kind that come with memorizing alphabetic lists -- a bad langauge-learning technique. I see no reason left for the complications and lengthening that this proposal requires (even ignoring that it scraps a pile of baselined material). Proper learning px>|83