Received: from spooler by stryx.demon.co.uk (Mercury/32 v2.01); 1 Dec 98 22:24:19 +0000 Return-path: Received: from punt-11.mail.demon.net (194.217.242.34) by stryx.demon.co.uk (Mercury/32 v2.01); 1 Dec 98 22:24:13 +0000 Received: from punt-1.mail.demon.net by mailstore for ia@stryx.demon.co.uk id 912494622:10:12520:0; Tue, 01 Dec 98 06:43:42 GMT Received: from pop.onelist.com ([209.207.164.13]) by punt-1.mail.demon.net id aa1122774; 1 Dec 98 6:43 GMT Received: (qmail 32723 invoked by alias); 1 Dec 1998 06:22:49 -0000 Received: (qmail 32659 invoked from network); 1 Dec 1998 06:22:47 -0000 Received: from unknown (HELO access5.digex.net) (205.197.245.196) by pop.onelist.com with SMTP; 1 Dec 1998 06:22:47 -0000 Received: (from lojbab@localhost) by access5.digex.net (8.8.4/8.8.4) id BAA26995 for lojban@onelist.com; Tue, 1 Dec 1998 01:19:21 -0500 (EST) Date: Tue, 1 Dec 1998 01:19:21 -0500 (EST) From: Logical Language Group Message-Id: <199812010619.BAA26995@access5.digex.net> To: lojban@onelist.com Mailing-List: list lojban@onelist.com; contact lojban-owner@onelist.com Delivered-To: mailing list lojban@onelist.com Precedence: bulk List-Unsubscribe: Mime-Version: 1.0 Subject: [lojban] UNL, Lojban and MLHT X-PMFLAGS: 34078848 7 1 Y0347F.CNM Content-Length: 1859 Lines: 42 From: Logical Language Group >On Tue, 24 Nov 1998, John Cowan wrote: > >> Sounds pretty crude and primitive compared to Lojban. > >Why not outperform the UNL vaporware then? >Could a Lojban parser output intelligible English some day? The glosser that nora has been working on will output somewhat intelligible English, and it is hardly as sophisticated as anyuthing the UNL people are contemplating. The problem is not the Lojban so much as it is making the English intelligible. No doubt an approach similar to whjat you describe of converting parse to Prolog to Lisp to English would be more effective than what Nora is doing, but Nora's solution is easy (if she only had time to work on it) A good glosser is a useful learning tool. A good translator would also be good. But the main thing needed to make either very useful is some kind of AI based error correction - something that couild look at a bad parse and figure out what was likely to be wrong (a missing "cu", two selbri in the bridi, etc.) and perhaps even suggest the corrections. Right now if a user makes an error that leads to non parsing the glosser cannot do anything with it, which in turn means that the user/learner has very little clue as to what was wrong. The YACC-based diagnostic that the parser puts out is useful only to those really aware of how a YACC parser works, and then only sometimes. lojbab ------------------------------------------------------------------------ Help support ONElist, while generating interest in your product or service. ONElist has a variety of advertising packages. Visit http://www.onelist.com/advert.html for more information. ------------------------------------------------------------------------ To unsubscribe, send mail to lojban-unsubscribe@onelist.com