From rpc@digitalkingdom.org Tue Apr 02 13:59:19 2002 Return-Path: X-Sender: richard@rrbcurnow.freeuk.com X-Apparently-To: lojban@yahoogroups.com Received: (EGP: mail-8_0_3_1); 2 Apr 2002 21:59:18 -0000 Received: (qmail 8971 invoked from network); 2 Apr 2002 21:59:14 -0000 Received: from unknown (66.218.66.218) by m12.grp.scd.yahoo.com with QMQP; 2 Apr 2002 21:59:14 -0000 Received: from unknown (HELO scrabble.freeuk.net) (212.126.144.6) by mta3.grp.scd.yahoo.com with SMTP; 2 Apr 2002 21:59:14 -0000 Received: from du-010-0052.freeuk.com ([212.126.153.52] helo=rrbcurnow.freeuk.com) by scrabble.freeuk.net with esmtp (Exim 3.33 #1) id 16sWJ7-0007GS-00 for lojban@yahoogroups.com; Tue, 02 Apr 2002 22:59:13 +0100 Received: from richard by rrbcurnow.freeuk.com with local (Exim 3.34 #1) id 16sWFT-000058-00; Tue, 02 Apr 2002 22:55:27 +0100 Date: Tue, 2 Apr 2002 22:55:27 +0100 To: lojban@yahoogroups.com Subject: Re: [lojban] ChomskyBot Message-ID: <20020402225527.A299@rrbcurnow.freeuk.com> Mail-Followup-To: lojban@yahoogroups.com References: <3C927952.20005@cs.technion.ac.il> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: ; from jay.kominek@colorado.edu on Fri, Mar 15, 2002 at 03:56:50PM -0700 X-Mailer: mutt-1.2.5.1i (Linux 2.2.19 i486) Sender: Richard Curnow From: Richard Curnow X-Yahoo-Group-Post: member; u=85274821 X-Yahoo-Profile: richard_p_curnow X-Yahoo-Message-Num: 13900 On Fri, Mar 15, 2002 at 03:56:50PM -0700, Jay Kominek wrote: > > On Sat, 16 Mar 2002, Alex Gontmakher wrote: > > > ChomskyBot, see > > http://www.helsinki.fi/~jppesone/applets/ChomskyBot.html, is an > > application that generates realistically-looking text from a given > > vocabulary (of course, missing any sense altogether). > > > > It would be interesting to apply the concept to Lojban ;) > > http://nuzban.wiw.org/wiki/?random%20sentence%20generator > mentions the LLG random sentence generator, and my random sentence > generator. > > Richard Curnow also has a system under development. > > (Mine use an algorith from a fellow out of .nz which analyzes the grammar > to determine all the proper weightings so you get sentences of exactly the > length you want. Nora's (the LLG one) just has hard coded weights. Not > sure what Richard's does, though I think he told me at one point.) Sorry for the late reply - I'm gradually reading through a huge backlog of Lojban list messages. Only a few hundred to go now! My 'algorithm' just has a lot of hardcoded weights and a few that are defined by variables that can be set at the top of the program. It's coded in perl. You can find it in the rtg/rtg.pl file in jbofihe's CVS module on digitalkingdom.org. To prevent endless recursion, a count is kept of how many times a rule has been picked that can 'go back up' the syntax tree. Once the counter hits a limit, only non-recursive rules are picked afterwards. It's all fairly Heath-Robinson. I only really wanted it for soak-testing the fix I did to jbofihe to cure some problems with the tense parsing that were found last autumn. Apologies for never releasing these fixes in a new version - I drifted onto some big updates to another project (chrony) around Xmas and haven't got back onto jbofihe since! -- Richard. ------------------+--------------------------------- Richard P. Curnow | Free software author + proponent Weston-super-Mare | a'o kansa ke skami se fanva United Kingdom | http://www.rrbcurnow.freeuk.com/