Return-Path: Message-Id: <9112081428.AA14227@relay1.UU.NET> Date: Sun Dec 8 12:09:50 1991 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: response to Cortesi on regular lujvo and glossaries X-To: lojban@cuvmb.cc.columbia.edu To: John Cowan , Eric Raymond , Eric Tiedemann Status: RO X-From-Space-Date: Sun Dec 8 12:09:50 1991 X-From-Space-Address: cbmvax!uunet!CUVMA.BITNET!LOJBAN Dave Cortesi writes: >The basic question is: should it be, or should it not be, a goal of >Lojban design that /le lojbo lifri tecusku/ can grasp the sense of a new >(well-formed) lujvo the first time, without need of more information >from /le cusku/ ? You've left out the important information, which is: is there context sufficient to inform of the sense. How do you learn new words in English? Mostly by absorption. You are reading along and digest the new word from context, sometimes not even realizing that you've never seen it before. Other times you have to stop and think about it a bit, and more rarely, look it up in the dictionary. But there are many quite fluent English readers/speakers that never look words up in dictionaries. Two examples: jabberwocky, where most of the words are 'new', yet the poem seems to make sense, (and James Joyce, etc.) gives a good idea of what the typical new Lojbanist goes through, except that the typical Lojbanist isn;t running into new words at quite so high a rate fo frequency. Methodology example: how you analyze to learn a Latinate root-based word meaning. You take the roots apart analyze their meaning generally from patterns you've learned for other words, then assimilate the compounds together. Anyone ever see the analyzed definition of supercalifragilisticexpialidocious (atoning for overabundant, educable, delicate beauty, I think the answer was). >Take the affirmative. It seems to me that this necessarily means that >lujvo must be restricted to a starkly regular pattern of arguments, >probably nothing more than the argument set of the terminal rafsi (same >as tanru). The reason is the mental burden on /le tecusku goi ko'a/. >By the time /le jvovla goi ko'e/ arrives, ko'a has already heard, >parsed, and stacked in short-term memory at least one, possibly more, >sumti. The use of new, coined words in speech will probably be proportionately much less than in writing, because the speaker has to do that too, in order to build the word, in addition to the time to think up the tanru basis in the first place. In real time, it isn't practical to do this all the time. You rely on patterns that have worked before, and been understood before. When you do make up nonce words, it is obvious by the pause in conversation, and if not, Lojban has the nonce word flag "za'e" to help the listener recognize this as a 'hard' word. Finally, in conversation, you have feedback: you can ask the speaker what the place structure meant. You usually will instead get a paraphrase, because explaining place structures in Lojban is probably harder to produce and to process than rephrasing. IN practice, though, I have usually found that in real-time speech, when I coin a nonce lujvo, I seldom have to do more than identify the tanru, and people can get the rest from context - learning the rafsi is far harder than learning however asystematically, to make understandable lujvo. IN written text, of course, you do not have the problem of overflowing stacks. you drop the stack and reread after figuring out what the word means. Finally, the assumption of starkly regular patterns of argument combining presumes something false as an assumption: that the typical Lojban speaker is going to know exactly the complete place structure for every word. I know less than 100 place structures well. I can guess accurately probably a few hundred more. I have made and plan to make no systematic attempt to learn them all. Instead, I rely on my ability, based on the English keywords that tell me what the word means, to guess on the fly what the place structures are for the gismu themselves. I don;t worry whether I am wrong unless I am presented with a sumti that makes no sense for the place structure I uhave guessed. If I am the listener, I try to figure out what a plausible place might be for an unidentifiable sumti given the source tanru. If I am the speaker, I choose my tanru so as to clearly include all the places needed to relate the sumti I wish to. The bottom line is that the more experienced Lojbanists are NOT having the problems you forsee. We don't make lujvo unless we expect them to be understood, either from simple pattern matching or sufficient context. More tricky lujvo, like the ones for mathematical terms, will probably not even have defined place structures until included in the dictionary. >It seems to me already a difficult mental task to unpack the rafsi of >ko'e and recall the argument pattern of the last one. If ko'a also has >to recall the argument patterns of the preceding one(s), and to assort >the pending sumti to different ones, and store also the mixed pattern in >short-term memory ready to receive the sumti yet to come in the bridi >-- ko'a is likely to drop all the balls. Only in speech, and the result is a simple "ke'o". Lojban conversations have a lot of "ke'o"s in them right now, with a moremajor reason being that the listener drops the balls on a complex sentence because the speaker hesitated too many times in composing the sentence that the listener is afraid that they forgot something. Quite typical when the speaker tries a sentence beyond his/her level of competence. >Yet as Nick noted, this severely constrains the possibilities of lujvo >for expanding the language's scope. Such restrictions would, and are impractical since people don't know all the place structures anyway. >Take the negative. It necessarily implies that >/le cusku poi zbasu le jvovla ku goi ko'i/ >has an obligation to provide that extra information. (I fault >Nick for not doing this in his recent Aesop postings.) That added information is usually context. YOU know that ants don't cause wheat, they might be able to cause dry-wheat from wet. YOu choose the meaning that makes sense. Now indeed, a computer will be rather less likely to understand LOjban with many nonce words - the language is much more ambiguous semantically. But it is still less ambiguous than English. People from other languages are able to learn that a railroad is not a road, but a train or system that uses rails as roads. I encourage people who make up nonce lujvo to post their proposed place structures. It is a good idea to do so, if people frequently misunderstand your lujvo, since it allows someone to get a handle on the types of things you are doing and to make suggestions (or at least to get through the material if it is the message that is important). But it is not incorrect or wrong or impolite not to. Several reasons. 1) Most important: I WANT people to write Lojban text and to post Lojban text, right or wrong. Any added work that has to be done in order to acceptably post material is liable to dissuade some people from posting. In Nick's case, he turns out massive chunks of text by Lojban standards, and nearly all of his lujvo are nonce lujvo. I will not ask him to provide interlinear translations, complete definitions of lujvo, or anything else. He is using the language, and his writings, if not perfect, are usually understandable. 2) Related to 1), and true, I believe, in Nick's case, is that not every Lojban learner is interested in teaching the language. A person interested in teaching should explore ways of getting new ideas across clearly. Someomne like Nick, who is trying to explore style questions (the neat thing about the Aesop postings is that they are direct from the original Greek, and Nick presumably tried to capture the Greek style. If you want easy material, you should avoid Nick's stuff. It will never be unless he chooses to make it so. 3) If you want the nonce lujvo displayed in English, some people would say this is rather impure an attempt to write Lojban. If you have to explain everything in English, Lojban isn't really a language, but a code. If it IS a language, then learning it will be subject to the same problems as every other language - no two people will have the identical vocabulary and dictionary of meanings in their minds. 4) If you want the nonce lujvo explained in Lojban, you must either choose a fancy convention, or for most people, the definition will be more complex than the original context. pc has formally attempted to define a couple of gismu totally within Lojban. He succeeded, but only Nora and I could probably have understood what he wrote even with a dictionary in hand. I've done the same thing several times in conversation when I don't want to lapse into English. But even then, neither pc nor I try to define the place structures in Lojban - the concept is hard enough to get across. 5. In any moderately complex text, the definitions will be longer than the text they presumably aid. If you've looked at the typical bilingual dictionary, it does not give definitions for the words - it gives an equivalent word. You have to use this information with disgression, of course - the grammar of the word in the other language may be such that though the translation match, the usage does not. In a good translation/student's dictionary, the definition that matters is typically very similar to a bilingual dictionary, but with example sentences and maybe some brief disambiguating notes - NOT elaborate explanations in either language. I would be surprised to ever see a Lojban English dictionary do even that much - space and time to create being factors. I see it as more important to have more words with briefer definitions than fewer words with comprehensive definitions. If the Lojban dictionary is 800 pages long and has only 5000 Lojban words in it, it isn't doing its primary job. >Someday there should be a dictionary of accepted lujvo; ko'i need not >define any word in it. Why would the existence of a dictionary matter? The listener isn't going to stop and look up the word in the dictionary. You want to see those balls in the stack drop: just watch what happens when a Lojbanist in conversation spends more than a small percentage of the time looking up words in the gismu list. Looking is a dictionary would be even slower. Furthermore, do you think that the speaker is going to look every predefined word up in the dictionary and/or use it correctly every time? Do you, a fluent English speaker, use every English word you know correctly without fail. You, Dave, as a professional writer - do you look up even a significant portion of non-technical words when you write? I find it much harder to deal in Lojban with people's 'errors' in gismu, than I do with errors in lujvo making or interpretation. The only gismu I urge people to concentrate on the place structures of, are those where two places are sufficiently similar in typical content as to likely cause confusion. Thus it is important to know that in klama, the destination comes before the origin, that with cinri, the x1 is the interesting thing and the x2 is the observer, with cfipu, that the x1 is the confusing thing and the x2 is the observer. fanva (translate) is the worst of all, because the old baseline gismu list does not have an adequate place structure for using the word, so NO ONE follows the standard, and indeed I know that I've caught myself using two or three different place structures for the word in conversation. But yet, noone has ever failed to understand one of my fanva sentences that I know of, because I ALWAYS make sure that there is enough additional context. (Providing context is not always an option, if le cusku is translating rather than composing. Aesop wrote very sparsely and vaguely, according to Nick. Unless he wants to produce an annotated translation or a paraphrase, there WILL be difficult passages.) A question, Dave. What do you see and expect from a non-fluent English spekaer who tries to talk to you? Does such a speaker even noticeably look up words in a dictionary when talking to you? What do you do, when such a speaker uses a word incorrectly by your knowledge? You ask for an explanation ONLY if you think you didn't understand. And I'm sure you can cope with the person who writes English and uses words incorrectly, because most people write abysmally even when they are fluent in speaking. If you make the burden on a learning, non-fluent, Lojbanist, more difficult than you require of an English speaker speaking English, your expectations will drive everyone, including yourself, from learning the language. If you are talking about written text, you have a different problem than with speech - no stack to worry about because you can reread as needed. But with Nick's stuff, for example, I am still going to try to guess what he means from context rather than from systematic rules. >But to use a new-coined lujvo without supplying >a gloss should be at least bad manners, since it shows that ko'i is >speaking without regard to whether ko'a understands. Actually, you could also think of it as a compliment. "ko'a" thinks that "ko'i" is bright enough to sufficiently determine the intended meaning from context plus the clues of the analytical source metaphor. >Other problems impend: if the lujvo argument pattern is the choice of >ko'i, different writers will inevitably coin the same lujvo with >different meanings and arguments. It's all very well to say that "usage >will determine which is best"; there will still be a residue of old >writings that use the non-surviving definitions. These will become >increasingly inaccessible with time. This seems a shameful waste, but >the only way to prevent it is to fix the allowed lujvo constructions >early. Ever read Chaucer or Shakespeare? This will happen no matter what we do, because: 1) gismu place structures will evolve 2) the nature of concepts will change, such that even knowing the place structure might give you a misreading on the concept and its connotations. 3) words will pass out of use and no longer be found in current dictionaries, then 100 years later may reappear with a new meaning. This is actually more likely in a constrained compound language like Lojban than in a langauge like English. 4) Moist important: people will make mistakes in place strctures even of well-defined words. Nick has reported that Zamenhof's Esperanto is now seen as archaic - he uses words and phrasings that are no longer common or accepted. He says that the first 'great' Esperanto poet (Kobe?) wrote stuff that abounds in word and grammar errors, but this has not diminished his stature. >What is worse, there could be contending usages that end in a draw, so >that there come to be entrenched stylistic "schools" of lujvo usage. >(yes, even despite the charitable, unegotistical attitudes so prevalent >in the world of conlangs... :-) This is only likely in a conlang that presumes that rules are supposed to govern lujvo. I oppose dikyjvo PRECISELY because I don't want them set up as a standard by which other approaches are 'wrong'. 'Schools' require rules. (I understand that Esp-o does have such schools, by the way, yet it doesn't seem to hurt the language. Indeed, doesn't English have MANY schools of usage and style for prose and poetry alike? Even at the level of word meanings and correctness (use of slang is verboten for some, and then we can always get into differences between British and American and ... 'schools' of English. Want a lift?) If two schools become totally implacable, you end up with two different dialects, with each stigmatized in the eyes of the other. We have Black English and 'Standard American' English, as well as Serbian, and Croatian in the news these days. The standardization force is precisely the desire to be able to communicate between schools. If school A and B are so implacable about the meaning of a word, then I suspect that neither will use it when they SINCERELY WANT to communicate with each other. Only if the two schools don't WANT to communicate with each other is the split impossible to resolve. Slangs and cants fall into this category. A teenager doesn't really want his parents to understand the 'in' dialect. In short, what you fear is a standard feature of natural languages. It cannot be avoided, and it is a waste of time and effort to try to. Lojban offers plenty of real advantages in disambiguity of grammar, analytical metaphors, etc, to provide significant communicative advantage over English. A second standardizing force will be the dictionary writers. Lexicographers try to be descriptive rather than prescriptive, but two things interfere. 1) There is never enough space in the dictionary to cover all the nuances, or all the words. The choice of what to leave out stigmatizes to some extent the words and meanings that are left out. This is heightened by 2) the fact that most people treat a dictionary as prescriptive even though no one writes dictionaries to be prescriptive. I think, though, that you will find far fewer areas of disagreement in Lojban word formation than in English word coining, and English is doing quite well, thank you. >Return to the matter of the obligation on ko'i (remember him?) to >provide a glossary. It seems to me an interesting question, whether a >lujvo can be defined _in Lojban_. Has anyone every attempted to write a >definition of any word, in Lojban? If the free-lujvo policy is adopted, >a conventional form for a glossary in Lojban is needed ASAP. Yes. Discussed above. There have been several varieties of glossaries, and we have explicitly provided two totally different ways within the language (ta'u, and cei) that will deal certain desired features of a glossary. >There is a middle ground, I suppose the one that jimc advocates: a >larger set of standard lujvo patterns, with pragmatic exceptions. Yet >he mentions "350" lujvo argument patterns, or cases? Good grief. My >immediate(*) reaction to that was, > >ganai mi djica le zu'o mi tadni le du'e nadikni valsi [kei?] gi me tadni >le fasban You made three errors in words, but I understood it just fine (naldikni, or na dikni; mi vs me and fasyban). The kei is not needed because the gi terminates all constructs back to the ga that it closes. Note that with Lojban logical conditionals, this sentence makes no implication at all about whether you want to study anything other than excessive irregular words. Of course, no Lojban words are irregular in the sense that French words are, so if you want to study excess irregular words, you don't want to study Lojban. The degree of regularity in Lojban words probably exceeds that of any natural language and most if not all conlangs (including Esperanto) MERELY on the strength of the uniquely resolvable rafsi. However, this brings to mind an issue people are ignoring. At least with lujvo you have a clue on place structures. If you are so afraid to guess (.iicai) a word's meaning and place structure, what are you going to do when you run into a le'avla borrowing. What is the place structure of 'djarspageti'? I want to note before closing that, even with the most regular of rules, fully 90% of all Lojbanists will rarely if ever coin a lujvo. It is the poets who add words to English. This will be true with Lojban as well - most people will limit themselves to what they can look up in dictionaries - except that a higher percentage of Lojbanists may be pioneering spirits that will dare to risk saying something that might be misunderstood, on the chance that they might just manage to be truly understood. That is the risk of poetry, after all. The problems Dave and jimc ask about are real, but primarily so because we have no dictionary, no fluent speakers to set authoritative examples, no established processes and algorithms that make learning a language seem simpler than it really is while obscuring real understanding. If you don't have the courage of the poet, it is probably too soon for you to learn Lojban just yet. But I have no doubt that anyone reading who contemplates learning Lojban is perfectly capable of successfully being a poet if they are willing to work at it, and are willing to be wrong quite often until they have the tricks down intuitively rather than algorithmically. lojbab