Return-Path: Message-Id: Date: Fri, 12 Jul 91 05:25 EDT From: lojbab (Bob LeChevalier) To: lojban-list Subject: response to Mark Shoulson on optional "cu" Status: RO X-From-Space-Date: Fri Jul 12 05:26:14 1991 X-From-Space-Address: lojbab Mark's confusion here may explain why we have to work hard on explaining how elision works. His questions are NOT stupid, since many people ask them. The change from 'elidable' to 'optional' for "cu" actually has no real effect on the grammar of valid Lojban sentences. It does not produce a shift-reduce error. We do not accept shift/reduce errors in our unambiguity proof based on YACC. But the change IS really tied into the elidable system, and that is where Mark gets confused. All of his problem examples do not parse as he expects, because he omitted critical elidables. I am responding on-line rather than with off-line preparation, so I cannot fiddle with his actual examples. So pardon if I use my own. is two sumti with no selbri. Optional 'cu' has no direct relevance here. It will not be inserted by the parser, which accepts a collection of sumti with no selbri as grammatical (a list of answers to ma questions, for example) What the parser WILL do is add in elidable 'ku' at the end of each "le" description: (le mlatu citka KU> VAU Because 'cu' is an optional instead of an elidable, it will never be inserted using the YACC error processing. It does not need to be, because every sentence that uses "cu" has an alternative form using some other elidable. Modifying the above example to one that makes the probably intended sense: le mlatu cu citka le ratcu The cat eats the rat. parses correctly, with the "cu" identifying that there was an elidable 'ku' after 'mlatu': cu So the 'real sentence' is le mlatu ku cu citka le ratcu ku vau IN THIS SENTENCE, THE "cu" IS 'OPTIONAL'. It is NOT possible to talk about optionality in a sentence where you have already elided, because taking the option changes the conditions for elision. Thus le mlatu ku citka le ratcu ku vau is equally grammatical to the version with "cu" in it, and indeed the final 'ku' and 'vau' are elidable regardless of the "cu", so le mlatu ku citka le ratcu works as well as le mlatu ku cu citka le ratcu cu and in this pair of examples it can be seen that the "cu" is 'optional', as opposed to 'elidable', since it parses the same with or without the "cu". Without the "cu" however, the "ku" after mlatu is NOT elidable, since the x1 sumti then combines with the selbri to form the 2-sumti phrase that I started with. The reason for making the grammar change is simply that the parser never really needs to insert an elided "cu". If it gets to the selbri, it must already have terminated any preceding sumti with their proper terminators in order to realize that there is a selbri to recognize. In such a case it does not need the "cu". On the other hand, HUMAN BEINGS prefer the "cu" to the rats nest of elidables. They know that if they insert the optional "cu", that they don;t need to worry about terminating a variety of constructs that might be seen before the selbri. Many examples of this are shown in draft textbook lesson 4/4A, though probably not enough or well-explained, since Mark got confused. HUMAN listeners also prefer the "cu" becausae it helps them, who parse more than one token at a time in their minds, to divide the sentence they hear in two - the pre-selbri part and the selbri+post-selbri part. In moderately complex sentence, for example with an abstraction clause having sub arguments in front of the selbri, this makes Lojban understandable: le nu le prenu ku klama le zarci kei ku se pluka mi The event the person goes to the store pleases me Uses 3 elidables and you may have trouble finding the two selbri (main and abstract) in the text. But it parses grammatically (unless I made a screw-up) and effectively identically to the easier to see version using "cu": le nu le prenu cu klama le zarci cu se pluka mi VAU) KEI} KU> cu with the "cu", two elidables before 'se pluka' became really elidable. Without the "cu" you need at least 2 of the 4 that were elided, and if they are the wrong two, you still get in trouble (you must have the second "ku" as one of the at-least-two , or without the "cu", the "se pluka" becomes part of a tanru of x1 of the form abbreviated as: le The key thing to remember is that elidables are NOT always elidable, they are OPTIONALLY elidable under specific if not well-stated conditions. This makes OPTIONAL 'cu' seem not quite so different from OPTIONALLY elidable 'cu'. The peril is that a description such as the grammar change report presumes that the reader knows how each of the constructs being changed works before and after the change. In this case, what really happened was that a path that the parser never needs to take - inserting an 'elided' "cu" - has been eliminated from the parser tables. In some earlier versions of the grammar, we had not been able to do this. Presumably the grammar is now stable and well-formed enough to allow the optional "cu". The disadvantage of the change,of course, is that with the "cu" not listed in the elidables, when the parser processes text, it inserts all the other elidables and not the "cu", and reading the parser output for a large block of text is more difficult without that large-print CU to tell you where the selbri is. This however is a small tradeoff for a cleaner and more honest grammar, and I still hope to eventually have a post processor that can automatically process parser outputs into some kind of structured readble form. But not this month. lojbab