[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Crosspost from comp.ai.nat-lang 2 of 2
The following is part 2 of 2: postings about Lojban that have been on
comp.ai.nat-lang in the last few days
lojbab
|From: holmes@diamond.idbsu.edu (Randall Holmes)
|Subject: Re: The artificial language Loglan
|Date: Sun Jun 19 14:03:04 EDT 1994
|Organization: Boise State University Math Dept.
|In-reply-to: alelyuna@brahms.udel.edu's message of 19 Jun 1994 15:07:04 -0400
|
|In article <2u250o$mpl@brahms.udel.edu> alelyuna@brahms.udel.edu (Robert Alelyun
|as) writes:
|
| Randall Holmes <holmes@diamond.idbsu.edu> wrote:
| [snip]
| >This seems like a natural venue to find people who know about it or who
| >might be interested in finding out about it.
| >
| >Loglan is an artificial language with a YACC-parsable grammar which is
| >intended to be spoken by human beings (but is (since parsable) a
| >possible means for communicating with machines). Its grammar is based
| >(not quite perfectly) on first-order predicate logic. My impression
| >is that it _is_ potentially a member of the genus "natural language"
| >(my impression is that people can and it is claimed that some have
| >learned to speak it); in fact, it is relatively easy to learn
| [snip]
| >
| >It has moderately good software support; a parser (from the level of
| >phonology up to the level of sentence structure) and an on-line
| >dictionary. The latter is a new product which makes life much easier.
| [snip]
|
| Quite interesting. I remember a previous post about Lojban, and its potentia
|l
| for improved natural language processing. I think this points out a popular
| misconception about what language is, and what makes it hard to process.
| These comments also apply to "limited vocabulary" approaches, such as
| restricting English to 500 words.
|
| My opinion is that the complexity of language comes from the enormity of the
| class of things which can be _meant_, not the class of what can be said.
| The class of things which can be _meant_ derives from the complexity of
| culture. It sounds simple, but it is a 180 degree turn in viewpoint from the
| a
| priori assumption of many modern attempts at language. Meaning flows from
| culture to language, not the other way around.
|
|See the last note. A lot can be said in Loglan/Lojban, but mostly in
|vague non-logically analyzable ways as in English or other NL's; there
|is no attempt to analyze meaning logically in these languages. The
|logical machinery implements first-order predicate calculus and possible
|some very simple set theory, but the predicate vocabulary is not
|logically systematic; it is built using metaphors which are not subject
|to logical transformations.
|
|
| What would this mean for artificial languages? It means that people will wan
|t
| to express the same complex communications, with the nuances and the shading,
| they are used to. If the means to do that is removed from traditional
| carriers of this information, then other carriers will be extended to do the
| duty. Once English itself (and I'm sure other modern languages) had a
| limited vocabulary, and a very simple grammatical structure (thematic roles
| were marked directly by word inflections, you can't get much easier to parse
| than that). Descriptions were taken in the concrete, as examples of single
| events placed in time, not the abstractions we are so facile with today.
|
| Consider a piece of language like:
|
| [In reference to Vikings]
| "Their ships were long and narrow... On the sails were painted devices like
|
| the eagle or the wolf."
|
| This is really quite odd, a special use of language for a special purpose.
| Note there is not a real ship discussed here, nor are we talking about "all"
| ships in the sense of logical ALL, perhaps we might use the modern notion
| of prototype to describe what this language is doing.
|
|Loglan actually happens to have a form of reference (a variant of the
|definite article) to handle this exact kind of generalization (reference
|to a "typical" object of a class). But I'm sure that other frustrating
|examples could be found!
|
| The point is, because
| people like to think this way, in terms of a concretized abstraction, the
| language allows the interpretation which should be obvious to all of us.
| Similar language can allow a strictly concrete interpretation, like:
|
| [In reference to Davis, a Viking]
| "His ship was long and narrow... On the sail was painted a wolf."
|
| My guess is, although I haven't studied this, is that the concrete language
| was first historically, and because people wanted to express themselves this
| other way, they extended the concrete to abstract under certain situations
| by conventional usage.
|
|Of course, analysis of "on his sail was painted a wolf" reveals some
|logical complexity!
|
| Then back to Lojban, even if a simple, parsable set of supposed meanings
| and rules is created, people are going to want to do the same complex
| things with the language, and will simply rely upon stricter contextual
| conventions; the net language processing task is equally difficult.
| You've just made one step shorter-- mapping from sentence to some logical
| form-- and made the truly difficult step longer-- mapping from logical
| form to purpose.
|
|Loglan or Lojban (two closely related but different languages) do not in
|fact restrict meanings; they share devices which allow the expression of
|the same kinds of vague modifications available in English or other
|NL's, and these will _firmly resist_ logical analysis (as opposed to
|parsing)! The only advantage (and it is noticeable) which the
|constructed languages have is that a genuinely logically analyzable
|sentence can be identified as such; in English it is easy to come up
|with sentences which look as if they have a certain logical structure
|and do not have it.
|
| The only really successful ways so far demonstrated in simplifying NLP
| involve restricting _domains_, which is actually a way to restrict possible
| purposes of the language. So my guess is either Lojban will be
| frustrating and inferior to real languages, or equally hard to process.
|
|My suspicion is that Loglan/Lojban would be usable in NLP for work on
|restricted domains; the advantage they would have is that they come with
|built-in parsing capability (already implemented) and, in principle,
|mechanizable recognition of valid logical transformations (I'm not
|talking through my hat here: I write theorem proving software, and I'm
|quite certain that Loglan/Lojban are suitable languages; my problem with
|them is that they are just slightly larger in their grammar than the
|sort of language I would really like to work with -- precisely because
|they contain grammatical arrangements for much non-logical stuff found
|in NL's!) The disadvantage is that one really does have to learn a
|language! Contra the last point, it might be easier to learn a small
|alien language than a grammatically constrained small fragment of one's
|own, for some?
|
|I agree with you; but Loglan does not claim to have a logical analysis
|of meaning; it has logic (the usual kind) built in, along with
|_non-logical_ language constructions capable of expressing nuances of
|meaning but resistant to logical analysis, and for both kinds of
|constructions is syntactically completely unambiguous (this is verified
|mechanically using compiler construction tools!) Loglan has unambiguous
|syntax, and the part of the language which is logical is capable of
|mechanically supported/verified transformations. But this does not
|scratch the problem of semantics!!!
|
| So those are two very different views-- one is to think of meaning as
| flowing outward from language into the culture, the other is to think of
| meaning as flowing from culture into language, that is, the language
| expresses exactly those distinctions people within the culture find
| relevant.
|
|Loglan was actually designed by a (logically competent) sociologist to
|address the Sapir-Whorf hypothesis, so its design does address these
|kinds of issues, but I don't know much about that side that side of it;
|I'm a logician and a latecomer.
|--
|The opinions expressed | --Sincerely,
|above are not the "official" | M. Randall Holmes
|opinions of any person | Math. Dept., Boise State Univ.
|or institution. | holmes@math.idbsu.edu
|
|
|
|From: helz@ecn.purdue.edu (Randall A Helzerman)
|Subject: Re: The artificial language Loglan
|Organization: Purdue University Engineering Computer Network
|Date: Mon Jun 20 01:52:11 EDT 1994
|
|In article <HOLMES.94Jun17211806@diamond.idbsu.edu>, holmes@diamond.idbsu.edu (R
|andall Holmes) writes:
||> Is anyone on this group familiar with the artificial language Loglan
||> (either Institute Loglan (the official version) or Lojban (the pirate
||> version); they are essentially the same, and I really have no animus
||> against the other flavor)?
|
|If you subscribe to the constructed languages mailing list
|(conlang@buphy.bu.edu ) and ask you can have all of your questions about
|lojban answered. The loglanites are not represented per se.
|
|I looked into using lojban as a basis for doing knowledge representation
|in NLP, and came to the (very personal and non-universally binding)
|conclusion that (for me, on my work, at the present moment) lojban
|really didn't buy me anything more than simply using first-order
|predicate calculus for the knowledge representation language, and lojban
|had the additional drawbacks of having a more complex syntax and a wierd
|vocab to memorize.
|
|However, its a "cool as hell" language and imho should be more widely
|known. The main lessons which the NLP and AI community can (in my view)
|draw from it are the following:
|
|1. How do you logically break the world into chunks and then assign
| words to each chunk?
|
|2. How do those chunks combine to refer to other chunks?
|
|3. How much of the world is reasonable to include in the language, and
| what should/must be left out?
|
|For me it was fascinating and instructive to see how some very bright
|people tackled these tough questions and how far they got.
|
|
|From: seth@wucs1.wustl.edu (Seth Golub)
|Subject: Re: The artificial language Loglan
|Date: Wed Jun 22 17:50:47 EDT 1994
|Organization: Washington University, St. Louis MO
|
|In article <CroLn0.BvF@noose.ecn.purdue.edu>,
|Randall A Helzerman <helz@ecn.purdue.edu> wrote:
|
|> However, its a "cool as hell" language and imho should be more
|> widely known.
|
|I agree. I've only recently begun studying it, but it has many
|features that I wish English had.
|
|There are a number of online resources for Lojban.
|
|The following is a reproduction of the Lojban contact info Web page.
|
|
|Getting into contact with la lojbangirz
|***************************************
|
|
|
|Write or call
|=============
|
| Bob LeChevalier
| The Logical Language Group, Inc.
| 2904 Beau Lane
| Fairfax, VA 22031
| (703) 385-0273
|
| e-mail: lojbab@access.digex.net
|
| o International Contact Addresses
| o Lojban Mailing List
| o Lojban FTP and WWW Archives Maintenance
| o Registration Form
| o Publications Order Form
|
|International Contact Addresses
|===============================
|
|Australia and New Zealand
| Nick Nicholas, nsn@vis.mu.oz.au
|
|Continental Europe and the British Isles
| Colin Fine, C.J.FINE@BRADFORD.AC.UK
|
|Scandinavian Countries and Finland
| Veijo Vilva, veion@xiron.pc.helsinki.fi
|
|
|
|Mailing List - lojban@cuvmb.cc.columbia.edu
|===========================================
|
|An active mailing list is available on the Internet. To join, send the
|message
|
| subscribe lojban firstname lastname
|
|to:
|
| listserv@cuvmb.cc.columbia.edu
|
|and send mailings to all mailing list members via:
|
| lojban@cuvmb.cc.columbia.edu
|
|Compuserve members can join this list, or can contact us, by preceding
|any of the above addresses with "INTERNET:". Fidonet connects with the
|Internet via a variety of nodes - contact your SYSOP.
|
|
|Maintenance of the Lojban FTP Archive
|=====================================
|( ra.cs.yale.edu )
|==================
|
| Erik Rauch, rauch@cs.yale.edu
|
|
|
|Maintenance of the Lojban WWW Archive
|=====================================
|( http://xiron.pc.helsinki.fi/lojban/ )
|=======================================
|
| Veijo Vilva, veion@xiron.pc.helsinki.fi
|
|--
|Seth Golub | "..And in all of Babylonia there was
|seth@cs.wustl.edu | wailing and gnashing of teeth, till the
|seth@hilco.com | prophets bade the multitudes get a grip
| | on themselves and shape up." - W. Allen
|
|
|
|From: delaques@gcg.com (Phillip Delaquess)
|Subject: Re: The artificial language Loglan
|Organization: Genetics Computer Group
|Date: Fri Jun 24 18:58:00 EDT 1994
|
|In article <2uabnn$no8@bigfoot.wustl.edu>, seth@wucs1.wustl.edu (Seth Golub) wri
|tes...
|>In article <CroLn0.BvF@noose.ecn.purdue.edu>,
|>Randall A Helzerman <helz@ecn.purdue.edu> wrote:
|>
|>> However, its a "cool as hell" language and imho should be more
|>> widely known.
|>
|>I agree. I've only recently begun studying it, but it has many
|>features that I wish English had.
|
|I agree too. Lojban is a fascinating new way to look at structure and
|meaning. I have been studying it ever since their WWW server was
|announced. Somewhere out there is the Lojban group's account of the
|split between Loglan and Lojban. I'd love to hear the other side of the
|story.
|
|===============================================================================
|Philip Delaquess | Ecx malgranda muso,
|delaques@gcg.com | Ne estas sen anuso.
|-------------------------------------------------------------------------------
|Bonvolu skribi min Esperante.
|===============================================================================