From jjllambias2000@yahoo.com.ar Sat Jan 22 07:12:47 2005 Received: with ECARTIS (v1.0.0; list lojban-list); Sat, 22 Jan 2005 07:12:48 -0800 (PST) Received: from web41907.mail.yahoo.com ([66.218.93.158]) by chain.digitalkingdom.org with smtp (Exim 4.34) id 1CsMwJ-0007ga-I5 for lojban-list@lojban.org; Sat, 22 Jan 2005 07:12:39 -0800 Received: (qmail 56263 invoked by uid 60001); 22 Jan 2005 15:12:08 -0000 Message-ID: <20050122151208.56261.qmail@web41907.mail.yahoo.com> Received: from [200.40.201.142] by web41907.mail.yahoo.com via HTTP; Sat, 22 Jan 2005 07:12:08 PST Date: Sat, 22 Jan 2005 07:12:08 -0800 (PST) From: Jorge "Llambías" Subject: [lojban] Re: Opinions, please: SA by structure To: lojban-list@lojban.org In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-archive-position: 9306 X-ecartis-version: Ecartis v1.0.0 Sender: lojban-list-bounce@lojban.org Errors-to: lojban-list-bounce@lojban.org X-original-sender: jjllambias2000@yahoo.com.ar Precedence: bulk Reply-to: lojban-list@lojban.org X-list: lojban-list --- "Adam D. Lopresto" wrote: > On the whole, I really, really like the idea. But I think a lot of things > need to be spelled out a bit more clearly. Questions I've got (there may be > others later): > > 1) Exactly what structures do we back into? I guess that's the heart of the > problem, and I know some discussion has gone on about it, but I think that > would be the hardest part to learn (as far as "no, you can't use {sa li'u}, > but you can use {sa lu}" (can you? Further thought seems to say "Yes, but it > goes back to the last sumti in general)). That's the idea. > 2) What happens if you use something that you can't (like {sa li'u})? Is it > simply a grammatical error, does it back out the entire discourse...? What's the difference? That you can't use any word after {sa} is nothing new, really. {i mi klama sa li'u} is ungrammatical with either scheme. > 3) What happens if you try to back out a construct that hasn't actually > occurred? {.i casnu sa mi}, for instance. That would erase everything, I would say. > 4) As a specific question, because my terminology isn't all that great, is a > sumtcita counted as part of a term? That is, does {mi tavla fo la .lojban. > sa > la .gliban.} keep the {fo} or not? (I'd be inclined to say not.) That's the idea. > Is {mi > tavla la .lojban. sa fo la .lojban.} legal? Yes. > 5) What happens with nested constructs? Does {mi kakne lo nu mi limna le > xamsi sa le lalxu} replace the {xamsi}, or the {lo nu li'o}? {le xamsi}. It's the one that stsrts last, as with {ri}. > Would a {sa > nelci} replace from {kakne} or {limna} (I think we've established it's not > {xamsi}.) {limna} > What about {le bruna be la .djan. sa mi}? {mi} would replace {la djan}. > 6) In general, using a structural approach instead of a strictly word-form > approach worries me in the presence of grammatical errors, and correcting > errors is what {sa} and friends are all about. I think fluent speakers will use {sa} not for correcting grammatical errors but for correcting semantic errors. A fluent speaker, almost by definition, does not make grammatical errors. I think the language should not be designed with beginners in mind. Begginners will make as many errors misusing SA as any other word, so it is pointless to provide such a complex way of correcting grammatical mistakes. The useful function od SA is for correcting semantic mistakes. > Specifically, I wonder > whether > you could confuse the parser enough that you couldn't get it to replace what > you want. {mi nelci le .djan. sa la .djan.} But *{le .djan.} isn't > grammatical, so will we be able to Do the Right Thing? In principle yes, because the parser will not look just for full terms, but also for anything that can be the beginning of a term. {le co pe poi sa lo} would also work. > And what, exactly, is > the right thing to do there? (Do we back out to the last *real* sumti, {mi}, > or do we understand the abortive attempt at a sumti?) That's debatable. My idea was to back out to any start of term. It has complications though. > I guess the last is my biggest concern. Everything else comes down to making > the right choices, but shouldn't be a fundamental problem. I just worry when > the way to fix grammatical errors relies on the grammar having worked > correctly up to that point. Not that I'm against it, but I'd like to hear > that it's being considered and that it's not really a problem because makau. I think the design should not focus on fixing grammatical errors, but I agree these things need to be considered. mu'o mi'e xorxes __________________________________ Do you Yahoo!? All your favorites on one personal page – Try My Yahoo! http://my.yahoo.com