Return-Path: Message-Id: Date: Fri, 5 Apr 91 05:56 EST From: lojbab (Bob LeChevalier) To: lojban-list Subject: GURP and NLP two great acronyms Status: RO X-From-Space-Date: Fri Apr 5 05:57:58 1991 X-From-Space-Address: lojbab a brief report on GURP (The Georgetown University Round Table of Linguistics, which Athelstan and I attended the last 4 days. There were just under 800 attendees. After initiially being hesitant for fear of adverse reaction from linguists, on Wednesday I put out about 30 brochures with a short note on Lojban's applicability to lingusitics research. They were gone within two hours. On Thursady I put out 110 more, and nearly all were gone when the conference ended at 4PM. We got some awesome name recognition out of this, even if none of these brochure readers decides to do something about Lojban now. Actually I suspect some will. Almost everyone we talked to seemed mildly interested in the concept of an AL designed for non international language purposes, and a coupole of researchers thought we had some interesting research angles that they might like to investigate. I would say that Athelstan and I together threw up more questions (usually good from the reaction of the audience and the speaker) than most people, so I'm sure we were noticed. There will be a lot of follow-up, and it won;t be high priority, since we have so many things going at once right now. But just one researcher might mean a share of a grant that would end our constant worry about money, as well as allow us to get Athelstan and maybe someone else working full time on the project, immensely speeding things up. So the postive reaction is heartening. The natural language processing people had some of the tougher questions for us. Jeff P. and Guy among the rest of you - I'd like some feedback and ideas since this is out of my expertise: - What are the relative advantages/disadvantages of using Lojban vs. PROLOG or LISP for internal langauge data storage and processing? Are the advantages of Lojban enough to justify someone spending the money to develop it? - Possible advantage mentioned for Lojban - easier construction and maintenance of a knowledge base written in Lojban, because it would be more "natural" than a programming language. - Another possible advantage - using a large corpus of English to Lojban translations coupled with a heuristic learning program to have the program 'learn' to translate English words naturally. Producing such a corpus would be relatively easy, since you add samples as needed whenever the computer makes a translation error. How tough is this, and has anything like it been tried? - How much work would it be to prepare a "Lojban compiler" that could translate Lojban sentences into a data base structured for efficient processing as is the data bases used by NLP that are loaded and accessed through LISP? - How much work would it be to create an English to Lojban 'translator' at the rudimentary level needed to show the feasibility, especially compared with existing parser/translator capabilities? - How much extra would it take for the heuristic processor I mentioned above? Some hard estimates would be useful here, as well as alternative ideas. I think a grant is possible within a year if we get good ideas and a credible approach. This list has some of the best and most experienced minds in Comp Sci., as well as some very creative people with perhaps less experience. I need help in this, and I've got some ears who will isten if I come up with good words that aren't pie-in-the-sky. -lojbab