Return-Path: Message-Id: <9205051916.AA00487@relay1.UU.NET> Date: Tue May 5 15:50:21 1992 Reply-To: cbmvax!uunet!viikki21.helsinki.fi!vilva Sender: Lojban list From: cbmvax!uunet!VIIKKI21.HELSINKI.FI!VILVA Subject: Re: syntax X-To: lojban@cuvmb.cc.columbia.edu To: John Cowan , Eric Raymond , Eric Tiedemann Status: RO X-Status: X-From-Space-Date: Tue May 5 15:50:21 1992 X-From-Space-Address: cbmvax!uunet!CUVMA.BITNET!LOJBAN > (Nick Nicholas, live from Whatchammy City) > Quoth Veijo Vilva >> {cripu ke fraso zbasu} >> {cripu je fraso zbasu} >> How do we interpret these phrases? >> {cripu ke fraso zbasu} > a bridge-related French builder >> {cripu je fraso zbasu} > a bridge-related and French builder >I'm not disputing this analysis (indeed, I sort of hinted at it >myself); I was rather pointing out the following: >a) the most obvious way a builder can be related to bridges is by >building them; >b) this interpretation is untenable in the above phrases, if you also >claim the builder is French. >If, however, you claim the relation between "French" and "builder" is >other than adjunct (for example, a builder of French things)... well, >even then I doubt you could claim the builder built bridges. >The problem I'm having is, how can {cripu} be an adjunct. At this >point, my Esperanto can help out; "ponta konstruanto" is not >equivalent to "pontkonstruanto". The semantics of a bridge as an >adjunct are still quite vague to me, though; the best I can see is >"bridge" being a locative. It ought to have been quite obvious to me from the beginning. There is a quite common situation where a bridge built is an adjunct: if we are not talking about bridges in general but a certain bridge, e.g. 'Here you see our new bridge. The bridge's French builder lives in our city.' The word "bridge's" is an adjunct which can be elided without changing either the meaning or the grammaticality of the sentence. (I think in Esperanto it would be quite legitimate to use "ponta konstruanto" in a case like this.) In this case 'cripu ke fraso zbasu' would be appropriate I think. If we elide the word "fraso" it might be advisable to retain the "ke". Then we should have: ripyzpa bridge builder cripu zpasu bridge builder or something-to-do-with-bridges builder cripu ke zbasu the bridge's builder In Finnish we have a quite similar situation: silta (gen. sillan) bridge rakentaja (gen. rakentajan) builder ranskalainen (gen. ranskalaisen) French sillanrakentaja a bridge builder sillan rakentaja the bridge's builder In the latter case 'sillan' is a genitive attribute defining the object of the building operation. ranskalainen sillanrakentaja a French bridge builder fraso ripyzba fraso ke cripu zbaso sillan ranskalainen rakentaja the bridge's French builder cripu ke fraso zbasu In this case in Finnish it wouldn't be correct to swap the words 'sillan' and 'ranskalainen' unless someone wanted to put a very strong emphasis on the word 'sillan'. ranskalainen silta (gen. ranskalaisen sillan, note the concord) the French bridge ranskalaisen sillan rakentaja the French bridge's builder fraso cripu ke zbasu (Sadly, quite many of the younger generation cannot clearly differentiate between e.g. 'sillanrakentaja' and 'sillan rakentaja', perhaps under the growing influence of English. You see more and more examples where people just put words in a row thus completely destroying the meaning, especially in ads translated from English. There is even a school of thought which maintains that we ought to adopt the English way as the Finnish way is 'too' difficult for 'most' people to master properly - read: including the advocates themselves.) (Technical jargon in English also has, of course, strings of words which cannot be parsed in any sensible way. Languages like Finnish and German can usually avoid these with sometimes quite monstrous concatenations of words - artificially constructed examples having 100+ letters and real ones very often 20-30 letters in a single word.) In 'cripu je fraso zbasu' (which I should consider syntactically correct though I might insert "ke" for clarity => 'cripu je fraso ke zbasu') I feel that in most cases "fraso" would dictate the semantics : 'a (something-to-do-with-bridges and French) builder' as 'a builder of {bridges|the bridge} and French things' seems to me slightly far fetched. I don't think anyone would try to connect "the bridge" and "being French" (which in a sense would be even ungrammatical). In Finnish we can have tien- ja sillanrakentaja = dagjevripyzba a road-and-bridge builder tien ja sillan rakentaja = dargu je cripu [ke] zbasu the builder of the road and the bridge This is actually the only way in Finnish so the Lojban forms I have used would feel quite natural to us -- and to Japanese people who would perhaps prefer to keep the "ke" to remind of the "no" of the Japanese genitive. The question of the semantics of the tanru is quite complicated and my interpretation has been very strongly influenced by my native language. It is, however, an existing natural language with non-artificial parallel expression models and I see no reason not to take this model into account when we are creating guidelines for Lojban tanru pragmatics. (I am not, however, advocating the wholesale adoption of the Finnish practice as it would redefine the distinction between tanru and lujvo.) The distinction between complements and adjuncts isn't always very clear and different languages have different rules governing the surface structure even if we sometimes are able to say that the deep structure is similar or even identical. Even if 'cripu' can be used to fill the 2nd sumti place of 'zbasu' and so would normally be a complement, separating it from 'zbasu' in a tanru (e.g. with 'ke' or 'fraso') would allow a looser (adjunct) connection and thus widen the scope of possible useful expressions. It is impossible to define a grammar for a natural (or nearly enough so to be practical) language such that it can always definitely tell whether a sentence is 'grammatically correct' (Goedel). The law of diminishing returns dictates that you must draw the line somewhere when you try to account for various phenomena. In a constructed language the syntax can be very precisely defined but the question always remains as to which depth. There are many contradicting goals: the grammar ought to be manageable, allow enough expressive power, prune as many 'impossible' sentences as possible etc. Forbidding outright the combination of certain classes of words in certain ways (e.g. "rel-bridge and rel-French") is, of course, possible but probably not the correct way to proceed. Must a bridi like "mi vasxu loi gerku" be tagged grammatically incorrect or considered syntactically correct but semantically non-sense? Quite many grammarians would say its incorrect but I should be inclined to disagree. In Lojban, of course, it would be a question of the restrictions imposed upon the 2nd sumti place of 'vasxu', is it syntax or semantics? And if we require that x2 is a mass word, does it impose a restriction upon the nature of the mass involved? Or does the Lojban mass abstraction somehow sidestep the whole issue? Veijo