Return-Path: Message-Id: Date: Sat, 15 Jun 91 02:46 EDT From: lojbab (Bob LeChevalier) To: lojban-list Subject: Bob Chassells' text and sumti-raising Status: RO X-From-Space-Date: Sat Jun 15 02:47:03 1991 X-From-Space-Address: lojbab Bob Chassell wrote: >.i mi cuska lu genai le zekri prenu goi ko'a ge kalri rinka le stela tanxe > I said "If the thief opened the lock box > > gi porpi rinka ri > and did not break it, (1) > gi ko'a cu djano lo nu kalri sazru le tanxe vorme > then he knew the process of open operating the box door. and later revised this to: >Here is what I think is the correct solution using GEK and GIK: > > ganai le zekri prenu goi ko'a > only-if the crime person X1 > > ge kalri rinka le stela tanxe gi na'e porpi rinka ri > both open cause the lock box and other-than broken cause it (2) > gi ko'a cu djano lo nu kalri sazru le tanxe vorme > then he knows the process open operate the lock box. The correction is better. There are some minor nits which I will not bother with, and Nora identified about 4 problems in the text that would be serious enough to interfere with understanding, but on the whole, Nora and I agree with John Cowan's judgement that this was an excellent sample text. But I want to take a closer look at one part of the above, which Bob corrected without comment in the second version, but for which both versions fail. The text conveniently serves as an example of why sumti raising is needed in Lojban. In (1), Bob had: > gi porpi rinka ri > and did not break it, There is something obviously wrong here: the English has a negation that is not present in the Lojban. In (2), Bob made this: > gi na'e porpi rinka ri > and other-than broken cause it This has the negation, but incorrectly. With parentheses showing how the text groups, this will be clear. (People checking word lists will note that "porpi" is not the correct word for the English, but we'll get to that in a moment.) > gi rinka ri > and cause it Rather than saying that the thief did not break it, the text suggests that he might actually have fixed it, or in some other way had a causal effect on its being non-broken. The translation of the first version suggests this is not Bob's intent. The correct way to add the negation is to use "na", though given the heavy use of connectives in the sentence, which Bob says was stylistically important, it would have been better still to incorporate the negation in the connective: ginai porpi rinka ri (3) and-not broken cause it gi na porpi rinka ri and not broken cause it In both cases, the negation has scope over the entire clause: gi na and not (By the way, I never commented on Bob's short summary of negation a couple of weeks ago. It seemed to be good insofar as it went, but did not cover such tricky questions as the scope of negation. It was just such tricky questions, which were never completely answered for Institute Loglan, that we wrote the negation paper in order to answer.) Given this correction, let us look at what else is wrong. In (1), he translated "porpi" as "break", which is correct. It is the intransitive "x1 breaks". I think, given his modification in (2), that he might have wanted "spofu" "x1 is broken". In either case, however, the sentence fails for another reason. The role of 'it' makes no sense. I am assuming, by back counting, that "ri" refers to the lock. The place structure of all tanru is that of the final term, so what Bob has really said in (2) is: > gi na'e porpi rinka ri and other-than-breakingly causes the lock. a) The thief did not cause the lock, breakingly, nonbreakingly, or otherwise. What he presumably did was other-than-cause the lock to break. This might be expressed in Lojban as: gi na'e rinka le nu ri spofu/porpi (4) and other-than cause the-event it is-broken/breaks It would be very difficult to turn this into a proper tanru with the place structure Bob sought. The closest I can come uses the following transformations: gi na'e rinka co se spofu/porpi ri and other-than causer of-type be-broken of it gi na'eke se spofu/porpi be ri rinka and other-than: be-broken of it causer These are both sophisticated and very Lojbanic, and difficult for the English speaker to grasp. By comparison, neither Bob, John Cowan, or Nora noticed a problem with "ri" in the effect place of rinka - they knew what it meant, even though what he said was something else. The expansion (4) shows why. The effect place of rinka should be an abstraction, and we natural language speakers supply the obvious one when presented with "ri" and the tanru-supplied "porpi" (even though the porpi has no grammatical control over the interpretation of that place). What we all did was natural allow for sumti raising the "ri" from the abstract clause. With sumti raising, the clause in (2) makes perfect sense, and in effect Bob has given us a new use for tanru - to specify the predicate of the abstraction. The result, though semantically ambiguous, was perfectly clear to natural langauge speakers who naturally sumti-raise. I have argued against jimc that just such effects will occur. It is short and convenient to sumti-raise, and people LIKE vague-but-natural-seeming tanru. They do not and will not perform all the sophisticated analysis that jimc's computer will in analyzing his 'diklujvo' - they will just express and understand, and damn the computer. The rules of a language have to be sufficiently natural that people follow them almost habitually from the start. If jimc imposes all kinds of formal 'diklujvo' rules, and people break them and still understand each other, he will never convince them that they are wrong and he and his computer are right to misinterpret what they find easily understandable. They will simply say: fix the compiuter program so it understands our human language - we shouldn't have to change our language for the sake of his computer. My sumti-rasing proposal, though intended originally to solve another problem, solves the problem of Bob's (2) without forcing him to say (4). It requires him only to be cognizant that he might any time he expresses a sumti. actually be sumti-raising. If the place structure would normally require an abstract, just say what comes naturally, but mark the sumti as raised from an unspecified abstraction. I contend it is easier to get people to learn to put in "tu'a" then to be sufficiently self-analytical as to what they 'really mean' to devise (4). Instead, finally putting all the corrections and sumti raising together, you get either: from (3) ginai porpi rinka tu'a ri and-not breakingly causes it-to-be or from (4) gi na'eke porpi rinka tu'a ri and other-than: breakingly-causes it-to-be or gi porpi na'e rinka tu'a ri and breakingly-other-than-causes it-to-be Of course, I would still try to teach them to use (4) - I, like jimc, believe that specific abstraction bridi are not particularly difficult to learn to do, and the extra syllables are not many. But I want the fall back, in case I am worng, or just in case the Lojbani want to do things stylistically different than I want to, of them being able to express themselves natural-language-ly, making only the minimal concession of tu'a to preserve the explicit indication of the logical structure. lojbab