[infurl]

Hi folks, I'm new here having just joined yesterday. I've been doing my best to read through the forums and get a sense of what's here already, but still have a lot to catch up on. I am married, I live in Tasmania, and I have two grown up children and two small hairy dogs.

About my work, I've been a professional computer programmer for more than 30 years and have been fascinated by robotics and artificial intelligence for as far back as I can remember. I mainly work with SQL (PostgreSQL), Common Lisp (CLISP) and C (gcc) on Debian Linux but am also proficient with XML and related technologies such as HTML, XSLT and CSS. I've also got a substantial amount of experience with C++ but I don't use it any more, preferring to use Common Lisp for its power and C for its performance.

You can find out more about me from my various websites if you want, and ask any questions that you like and I will try to answer them.

ASMITH dot ID dot AU (my home page and my older robotics experiments)

INFURL dot NET (my recent robotics experiments)

WIXML dot NET (my computational linguistics database)

Any of you who are interested in natural language processing or computational linguistics may find some useful data at the wixml website. I've spent a great deal of time collecting and converting a variety of artificial intelligence resources into a unified relational database format and this work is ongoing.

Currently I'm developing a GLR parser in C and Common Lisp which can compile any context free grammar into a C program. The resulting code converts plain text into XML for further processing, or by using it as a C library, the parse graph can be accessed directly. Being a GLR parser it also handles ambiguity easily and efficiently. Still haven't decided if/when/how I might release this software, but at the very least, I hope to make it available using the "software as a service" model.

I'm also active on the CHATBOTS dot ORG website but as of last week I'll be conducting as much of my online business as possible through Google+ so I hope to see you there too.

Cheers,
Andrew Smith

[DaveMorton]

Andrew! Welcome! I'm shocked that it took you so long to get here! But I'm delighted that you've come.

Once you get a few posts under your belt, you'll be able to post URL's, so that we can have a peek at your sites. I think that three posts is the "magic Number" there, but I'm not 100% sure.

In the mean-time...

Ladies and Gentlemen, I'm pleased to introduce my good friend Andrew. He hails, not only from Down Under, but also from chatbots.org (which he's already mentioned, of course), where we've had some wonderful discussions. Please make him feel welcome, despite the fact that he knows me. That unfortunate fact should not  be held against him.

[Bragi]

Hi Andrew, Jan here. Nice to see you on this side of the web as well 

[infurl]

Getting a sense of deja vu but it seems I've only been seeing half the action all this time. Great to see you guys again. How many other familiar faces are here?

[DaveMorton]

Well, Roger Davies (AKA Freddy, here) is about. In fact, this is "his" forum site (though he insists that it belongs to us all), and there's the odd visit from time to time from Victor, though he hasn't been active since May. 8pla also stops in from occasionally, but that's about all that I can think of, off the top of my head. Others drift in on an infrequent basis, of course, and I'm sure there are others I haven't mentioned, for which I'm certain to be chastised over, at a later date.

[Art]

Welcome Andrew!! Always nice to accept a new arrival (even if you do know other certain "questionable" individuals whos name I shall not mention).

Looking forward to some enlightening discussions!

[Data]

Welcome to Ai Dreams Andrew

A professional computer programmer, well I confess you are on another level to me, I believe I fall into the "questionable” category, yeah that’s fair to say. 

Anyway here’s hoping you bunch of clever people that have accumulated here can come up with a new type of Ai, something that actually seems intelligent when you chat to it.

Its going to happen, I believe.   

[Freddy]

Welcome to the site Andrew and thanks for your introduction.  There's some very juicy stuff there

I'm trying to learn C# myself at the moment, I'm picking a lot up by looking at existing code.

I'm sure Datahopa is right.  Feel free to drop in any ideas you want to discuss, there's nearly always someone who has another angle or a fresh piece of insight.

I was looking at Google+ the other day funnily enough, when I clear some of my current projects I will take a better look.

[DaveMorton]

I was looking at Google+ the other day funnily enough, when I clear some of my current projects I will take a better look.

Well, when you do, be sure to look me up, as I'm there, as well. That goes for anyone else here that uses G+.

[Freddy]

I will as soon as they open it up.  At the moment they are saying they are full up.

[DaveMorton]

I think it's available by invitation only, Freddy. Fortunately, I think I can send out one or two invites. I'll send you one, if at all possible, yes?

[Freddy]

Thanks and done    Will take a look around later, bit busy at the moment.

[mendicott]

Greetings Andrew,

I'm now following you on Twitter at http://twitter.com/infurl ..

I hope you'll forgive me for saying so, but that's very Aussie of you (self-deprecating) not to include hot links to your interesting websites ..

http://asmith.id.au/ (home page and older robotics experiments)

http://infurl.net/ (recent robotics experiments)

http://wixml.net/ (computational linguistics database)

I think I can say this, being a quasi-Aussie myself, Australian-American ..

BTW, I had to look up GLR parser http://en.wikipedia.org/wiki/GLR_parser ..

In fact, I spent all day yesterday resuscitating my http://www.meta-guide.com/ project website ..

You may or may not have already gathered from by blog at http://www.mendicott.com/ that I've been trying to nut this out for some time ..

BTW, I believe Bruce Wilcox is also working on this at the moment ..

Richard Wallace does have a bespoke Pandorabots SpellBinder service for converting screen play dialogs into chatbot knowledgebases ..

The burning question is how to convert books (ePubs) into XML and RDFa, and then into natural language for creating "interactive talking books" ..

So far, I've determined that Adobe FrameMaker is an enterprise tool that converts books into "editorial XML", and TopBraid Composer can convert XML into RDFa ..

There are a number of people, including Ben Goertzel, Stephen Reed and David Burden, claiming to have conversational agents based on semantic web technologies, but I haven't seen any of these available publicly yet ..

I am definitely interested in any potential SaaS API, and look forward to learning more about your work ..

Cheers mate,

 - Marcus Endicott
   http://twitter.com/mendicot (note, new account with one "t")

[infurl]

Hi Marcus, I didn't post hotlinks to my websites because they are automatically deleted from new member's posts to prevent link spamming. After a certain minimum number of posts (three?) I should be able to post hot links. Thanks for the follow on Twitter but I am seriously considering deleting that account. Although I've met some good people on Twitter they're all using Google+ now too, and I'm finding even more people to follow using the tools available on that platform.

Parsing is one of those things that looks like it should be easy to do but isn't. Most programmers who write some code for parsing end up with something like a recursive descent design which gets the job done, albeit with lots of backtracking. These are very inefficient for more complicated grammars but their performance can be improved somewhat by using a technique called memoizing where the results of function calls are stored and re-used instead of being recomputed on every call. I believe that Victor was doing something like this at one stage.

A better approach is to build a chart parser which keeps track of all the different parsing branches and never does the same thing twice. This has much the same effect as a good memoizing implementation, but being coded explicitly for it, is much more efficient. Chart parsers can work top down (Earley Parser, named after the person who discovered it) or bottom up (CYK Algorithm, named by the initials of the people who discovered it).

The Earley parsing algorithm is pretty easy to understand and implement and you can have endless fun figuring out ways to make it go faster. It was during my experiments with an Earley parser that I wrote (first in Common Lisp, and then rewrote in C) that I found how much faster it would go if I merged the sub-trees representing ambiguous parses wherever possible. With a bit of research I found that this was in fact the essence of the Tomita GLR parsing algorithm which was regarded as the fastest parsing method of all.

So I discontinued my efforts with Earley parsing and set about learning all about shift-reduce parsing and graph structured stacks. Tomita's original 1985 paper is fairly brief and though it has comprehensive examples worked through in detail, it leaves much of the implementation practicalities up to the reader to figure out. This I duly did, again using Common Lisp and C.

A few days ago on Google+ I was told about a more recent paper about GLR parsing which improves on the original Tomita algorithm so I looked it up. This is called "Parser Generation for Interactive Environments" and was written by Jan Rekers and published in 1992. It has much more detail and is pretty easy to follow. Even better, the appendices contain the complete source code for all the software that was used in the research. Unfortunately it is written in an obsolete dialect of Lisp, but it doesn't look like it would be too difficult to convert to Common Lisp. Also, being Lisp, it is only a few pages of fairly straightforward code where it would be thousands of lines of gibberish in practically any other programming language.

Which leads me to ask, are there any other Common Lisp programmers frequenting this site?

Cheers,
Andrew Smith

[mendicott]

Andrew,

Good stuff, thanks!

This is all I've got for Lisp.

>> http://leafproject.org .. open source robotics project using artificial intelligence and vision .. #video #LISP <<

>> Probabilistic Learning Of Programs .. a Common Lisp framework for meta-optimizing semantic evolutionary search .. http://plop.googlecode.com <<

>> Mycin http://en.wikipedia.org/wiki/Mycin ..early expert system ..written in Lisp ..emphasized the use of judgmental rules that had elements of uncertainty <<

Cheers, Marcus