Return-Path: From: cbmvax!uunet!math.ucla.edu!jimc Return-Path: Message-Id: <9106112350.AA26579@euphemia.math.ucla.edu> To: lojban-list@snark.thyrsus.com Subject: Response to Six Messages Date: Tue, 11 Jun 91 16:50:46 +0100 X-From-Space-Date: Wed Jun 12 09:31:24 1991 X-From-Space-Address: cbmvax!uunet!math.ucla.edu!jimc Summary: 1. Seen one way, you need to "lower" sumti, not "raise" them. 2. The "sumti - brivla - sumti" phrase form is a major organizing structure for Lojban; the surface structures should avoid burying sumti inaccessibly inside abstract sumti. 3. The hidden predicates needed to interpret non-abstract sumti in abstract slots are outrageous. 4. A pan-predicatist solution to converted . 5. The problem with djuno "know" is that people put the wrong sumti in the wrong places. 6. The problem of cleft places has not gone away. (true, but not discussed.) (msg. 1) "Sumti raising" is what it's called in NL? And I have been assuming that the problem is that the sumti starts out at main level and has to be "lowered" (by copying, says I) into the subordinate clause. I have a very bad feeling about emphasis on "sumti raising". It sounds like you start with a surface structure with all sumti at main level. (we're talking about jufra but the same considerations apply to contained bridi too.) Now you (lojbab) assume that transformations magically occur which precipitate sumti into sub-phrases (abstract sumti or subordinate clauses) not said explicitly. Finally, the job of "sumti raising" is to restore the lowered sumti to main level. Is this really what you mean? I would start with the surface structure in which sumti are at the main level. Rules defined by the language architect specify transformations, in which sub-phrases are created and are populated with main level sumti. Thus the problem is to "lower", not "raise", the sumti. Various design choices can then be made about how to do these transformations. For example, I incline toward cleft places, such that the subphrase sumti are copies of main level sumti, whereas other people like uncleft sumti in which main level material is moved, not copied, to the subphrases. Similarly, I prefer very specific rules (cf. diklujvo) for what phrases are created and what material should be moved, whereas other people are willing to rely on the user's "coai quantity quantity of...on scale... le'a in category (scalar set) sele'a as a category of tele'a defined by quality (category) klesi les kle lei class class/category of...with property... li'e led by seli'e leading lidne li'e leader leader of...in purpose... ma'i in reference frame (by standard 2; cf ja'i) sema'i as a standard for (of reference) [??] manri mar reference reference frame for observing..w/rules.. mau exceeded by (a greater) semau more than (relative!) [??] semaunai not more than (relative!) [??] zmadu zad zma mau more exceeds...in property...by amount... me'a undercut by (a lesser) seme'a less than (relative!) [??] seme'anai not less than (relative!) [??] mleca lesser ...is less than ... [a mekso bridi] me'e with name (requires quote) seme'e as a name for (the named one) teme'e as named by cmene cme me'e name name of...to/used by... mu'i because of motive semu'i motive therefore [??] semu'inai motive nevertheless mu'inai despite motive mukti muk mu'i purpose purpose/goal for...(obj./event) 'motive' mu'u exemplified by semu'u as an example of mupli mup example example of... ni'i logically because (entailment) seni'i entails therefore seni'inai entails nevertheless (denied entail) [??] ni'inai despite logic (denied entail) nibli nib ni'i necessitate necessitate/entail/imply...for rules... pa'a in addition to (parallel) sepa'a paralleling (same as 1st) panra parallel parallel to...in geometry... pa'u having component sepa'u as a part of pagbu pag pau part part of whole... pi'o used by sepi'o using tool (instrumental) pilno pli use use...for... 'employ' po'i in the sequence (in order) sepo'i sequenced by rules porsi por poi sequence (pl.set) sequenced/ordered by rules... pu'a pleased by sepu'a in order to please (for) pluka puk pu'a please please...by doing/being... pu'e by process (in manner 1; cf ta'i, tai) sepu'e processing from (inputs) tepu'e processing into (outputs) vepu'e passing through stages pruce ruc ru'e process process: inputs..outputs..steps/stages.. ra'a pertained to by sera'a concerning (pertaining to) srana ra'a pertain pertains/relevant to... 'concerns' ra'i from source sera'i as an origin for krasi kra source source/start/beginning of... 'origin' ri'i experienced by seri'i experiencing lifri lif fri experience experience... 'life' rai with superlative serai superlative in (property) terai at extreme (superlative) verai superlative among traji taj rai superlative superlative in property... 'extreme' ri'a causal because (phys./mental) ri'anai despite cause (phys./mental) seri'a causal therefore (phys./mental) [??] seri'anai causal nevertheless (phys./mental) rinka rik ri'a cause cause for...under... sau requiring sesau necessary for (process) tesau necessarily under (conditions 2) sarcu sac necessary necessary for process...under... si'u aided by (aiding agent) sesi'u assisting in sidju sid dju help help/assimmon sense" to fill in ellipsized vacant places or to supply entire predicates. Or perhaps what you meant was, that people are supposed to explicitly say jufra (or s-bridi?) with abstract sumti arguments, but for various reasons the sumti buried in the abstract sumti have to appear at main level -- the problem addressed by -gua!spi retro-replication. Is this what you mean? Retro-replication is needed fairly rarely, but is a godsend when needed. (msg 2) It looks to me as though gasnu is properly the "generalized transitive" predicate which in Old Loglan was madzo/makso (from English "make"). For example: le mlatu cu gasnu le nu le ratcu cu morsi the cat (transitive) the rat dead The cat kills the rat (makes it dead) Ask your standard natural language speaker whether "kill" is a unitary concept that "ought" to be represented by a gismu-like brivla, and he certainly will say "yes, and the above pile of phrases is a pile of something else". In other words, a lujvo "mroygau" is demandest/aid...(obj. or event) 'aid' ta'i having form (like) (in manner 2; cf pu'e) seta'i as a form of tarmi tam tra shape shape/form of... 'form' tai by method (in manner 3; cf pu'e, ta'i) tai as a method for tadji method method/technique for doing...under... ti'i suggested by seti'i suggesting stidi tid ti'i suggest suggest/propose idea/action...to... ti'u associated with time (for letters) seti'u at the same time as (for letters) tcika time of day ..(hours,mins,secs) is time of day..at.. tu'i associated with site (for letters) setu'i as a location of stuzi tuz stu site site/place/location of... 'position' va'o under conditions (conditions 1; cf tesau) seva'o as conditions for (environment) vanbi vab environment environment/surroundings of... 'ambient' va'u contained by (container) seva'u containing vasru vas vau contain contains.../vessel containing...'vessel' zau approved by sezau approving zanru zar zau approve approve plan/action...(obj./event) zu'e with actor (purposed agent) sezu'e purposefully doing tesu'e with goal zukte zuk zu'e act act/do...with purpose/goal...; entity d. Jimc says that handcrafted lujvo are a dead end and proposes diklujvo -- a set of rules whereby virtually ALL lujvo and tanru, about 99% by count of usage in -gua!spi, can be interpreted to transform to a collection of phrases headed by gismu like the one above. diklujvo are much more attractive than requiring people to speak explicitly all the abstract sumti. (back to msg 1) NL's "I try the door" assumes a default predicate applying to doors, namely kalri "open", or sometimes ganlo "closed", depending on complicated context. As you say near (4), there is a lot of stuff hidden in there. The situation is reminiscent of the word "do" for which many objects have an associated behavior, e.g. to "do" a container means to move typical contents from x2 to x3 (via x4 the container? Let's not get sidetracked :-) If I get your drift, you ask "what does it mean when a non-abstract sumti turns up in an abstract slot"? You say, "hundreds of brivla include vague default predicates and the listener is supposed to drop the non-abstract sumti into x1 of the default predicate". I strongly recommend that such default predicates be discouraged. A predicate should mean THAT PREDICATE, not some other relation contextually defined via poetic license. The default predicates are too illogical! Particularly, major semantic surgery on the language cannot be justified by the (negative) utility of interpreting such non-abstract sumti. Also, it's common for someone to use a colorful non-abstract metaphor such as le to'ercnici "the mess" to refer to an event. How can you tell whether the speaker wants the default predicate engaged, or the event referent of the non-abstract metaphor to be used as such? More common sense? I see no option other than to take the user's words to mean just what he said. Cowan's (tu'a mi) galfi le bitmu, is being interpreted as (lenu mi cinta) cu galfi le bitmu. But this sentence again expands to (lenu mi preja loi cinta) cu galfi le bitmu. YECCH! All these "unspecified" predicates that get specified by imagination -- they bother me. (msg 2 again) I think you need (gau mi) cu galfi *se* le bitmu, because galfi x1 is now missing. I can accept this version much more than I can accept the other. You mention a distinction between the agent of change and the event that causes the change. I agree. Actually the current definition of galfi includes separate places for both. Here it is with my annotations: "(actor) x1 modifies (cleft object) x2 into (bridi about x2) x3 by (causative event about x1) x4" where "about" means that the bridi includes (a copy of) x2 or x1 as one of its sumti unless common sense ellipsis (or auto-replication) is blocked by explicit words. A comment from John Cowan: infinite precision requires infinite verbosity. However, by adjusting the rules you can make a given number of words be more or less precise. Many of my proposals have been to shift ambiguity and requirements for verbosity from one area of the language to another, thereby to make more precise the most common forms. (Msg 3) Alternative #1: Athelstan (?) objects to this line of evolution: gau mi galfi le bitmu (bridi) le jai gau galfi be *fi* le bitmu (similar sumti) However, this is bogus because the first sentence should have been gau mi galfi *fe* le bitmu (bridi should be) (since the original x1 is vacant). So in both situations the user has to jump over a place. The change from fe to fi is simply a consequence of shifting x1-x5 to x2-x6, presumably a tolerable change. Assuming you come up with a placelink for x6. Aside-note: In -gua!spi I found that exchange conversion (x1 becomes x4 and x4 becomes x1) was making trouble; a count of instances revealed that it was quite common to want to put a sumti in the original x1 which is now out in left field. Thus I changed over to what you call push-down conversion. Because the place numbering gets all wacky, I felt it was necessary for placelinks to refer to places before conversion, e.g. in this example "fa" would hit the original x1 even though it was then the second ordinal case (4th in Lojban). John Cowan told me, though, that in Lojban refers to places after conversion: "fe" or "fo" in this example. In any case, I probably like alternative #2 better because it messes up the place numbering less. Why bother to restrict JAI+tag from logical connectives, modals, etc? The meanings are totally freaky, but like you say, maybe Nick can come up with a valid use for them. Actually, as a pan-predicatist, I would say to use the gismu which the is derived from, with the original main bridi becoming an abstract sumti of it. Example: gau mi galfi fe le bitmu The wall was modified by me le gasnu be le nu galfi fe le bitmu The actor in the wall modification This usage is perfectly clear and it savages the language much less than a conversion for . To my way of thinking, the is actually an abbreviation for a predicate relation on gasnu, which is not approximately but exactly represented in converted form by the sumti shown. (msg 4) I find the discussion of djuno "know" together with relative questions to be rather opaque; I don't exactly see what the issue is. I agree that the definition of djuno should be something like "x1 knows (abstract sumti including x3) x2 about x3 (with epistomology or whatever) x4". I find for diklujvo production that it's more convenient to exchange x2 with x3: "x1 knows that x2 is (abstract) x3" but that decision comes a lot later. I also recognize that you might want to say "x1 knows (set of facts in extension) x2" where each fact is an instance of geometry. In English you would equivalently say "I know geometry" with "geometry" as x3. For logical expression, on the other hand, I would suggest something like this. One alternative is the Old Loglan word "spuro" meaning "competent", which I don't find an equivalent for in the gismu list. However, it has to be interpreted as "x1 is competent to do (abstract sumti containing x1)" and thus has the same semantic problems. What you really want to do is define "geometry" or other areas of knowledge or competence in this form: "(abstract sumti) x1 is a fact / action / etc. of (whatever field)". Thus: saske = "x1 is a scientific fact ...(possible other places)". Then "I know science" gets translated as mi djuno le saske I know something (cued by the word) scientific mi djuno so'a lo saske I know lots of scientific facts The events which are x1 occupants of saske are being correctly placed in djuno x2 which takes such events. (Of course competence in using the stuff is important too -- an opportunity for a parallel diklujvo. Also science is as much a process as a set of facts so the above definition needs tinkering, but the basic form will remain with an abstract sumti in x1.) Thus the difficulty is not in djuno, but in saske! I think the other examples are similar in that the problem is not with djuno, or with cleft / uncleft places, but with constructing the sumti used in djuno. (msg 6) An aside: in words for works of art, I have found it convenient to include a place for both the author and the performer. I had not previously appreciated that speaker, author and performer might need to have separate 's, but now I'm convinced. -- jimc