Return-Path: Message-Id: Date: Fri, 17 May 91 01:54 EDT From: lojbab (Bob LeChevalier) To: lojban-list Subject: gismu morphology - response to Chip S. Status: RO X-From-Space-Date: Fri May 17 01:54:41 1991 X-From-Space-Address: lojbab Chip Salzenberg , writes: > John Cowan writes: > >Mostly for better recognition. "blanu" = "blue" would be considerably > >less recognizable to English-speakers if it didn't end in "u". > I find this argument less than compelling when applied to a language > that is intended to be learnable by anyone. The Lojban words were made to maximize recognition scores based on 6 major languages. The result is words that are unfamilar but non-random. The actual benefit of the recognition scores has not been scientifically tested. > I forget the total number of lojban consonants. Assuming 18 > consonants and the current rules, there are about 60,000 possible > unique gismu. > > Yes, that's a lot (more than I expected, actually). But I am still > concerned that limiting the total number of gismu will eventually > cause a real problem. After all, we *will* run out someday. The number of basic roots in most languages, comparable to the gismu tends to average in the 500-5000 range. This does not count compounds and technical terminology. e.g. BASIC ENGLISH, a simplified English has around 1000 words. Lojban gismu are added very slowly, and nowadays have to be justified on the basis of their value in making new compounds. (Indeed, and I don't think it has been recently mentioned, our official policy is that if a word cannot be made using the recognition score algorithm, but justifies a gismu, then we assign it arbitrarily to a low density part of gismu space. Thes set of possible gismu is much smaller than 60000, because we do not allow certain voiced/non-voiced oppositions. There is also a very limited set of permissible initial clusters, and a less restricted set of medials. I think, for example, that there are 48x5x11 = 2680 possible CCVCV gismu and a substantially larger number -but same order of magnitude CVCCV gismu > If the trailing vowell is truly useful in distinguishing gismu, then > isn't dropping it during lujvo formation a bad idea? lujvo do have lower reundundancy than gismu. But you can always use the expanded tanru, or even longer 'unambiguous' expanasions if needed. You can't expand a gismu, though. lojbab