Post by KnyteTrypper on Oct 13, 2006 19:49:57 GMT -5
I hadn't looked for the transcripts, previously. That's maddening, lol. I wasn't able to find a normal transcript.
It's already commonly said in that circle "It's Hugh's contest, it's Hugh's rules." Unfortunately, it also seems to be as much about who is Hugh's favorite of the moment as otherwise. How else to explain why ALICE went from winner to "also-ran" at the same time Hugh became enraged at Dr. Rich for participating in "that other Turing test." Dr. Rich has been too much of a gentleman to even allude to such a thing himself, but I don't have a dog in that fight, so I have no qualms about pointing out what is at least the obvious appearance of the situation.
2006 is the first year where the Loebner transcripts aren't in standard, human-readable form. I don't understand why the real-time key-press typing simulation became so important in the contest. In human-to-human cellphone text messaging, instant messaging and chatrooms, input is typed first, then sent as a block of text. So if the aim is to simulate those situations, it doesn't.
Included on the transcripts page is a file called LoebnerPlayer.pl, and a rather complicated installation and run procedure, to "replay the transcripts in simulated real time." However, I can't do this on the library and borrowed computers I use.
It's interesting that you picked the word "maddening" to describe these raw-data transcripts, Knyte, because that's exactly the word that came to my mind. Interesting, too, that you pick the metaphor of a dogfight to apply to chat-robot contests. The image is male, violent and highly competitive. This is why I often shrink from public engagement.
Post by KnyteTrypper on Oct 14, 2006 23:22:39 GMT -5
nteresting, too, that you pick the metaphor of a dogfight to apply to chat-robot contests. The image is male, violent and highly competitive. This is why I often shrink from public engagement.
If you only knew, lol. The phrase derived from the image in my head which I gained from membership in the Robitron group, playground for the Loebner elite, where the giants of a.i. come to converse with one another. My original imagery, reading them snipping and sniping at each other like ill-mannered children, was that of a female canine and a conflict involving an exchange of open-handed blows. In deference to your known sensibilities I diluted it to "dogfight" imagery.
I think really all the fancy verbage about the "Loebner player" is really just saying the pc you use has to have the Perl library installed. That's not an entirely unreasonable thing to expect from even a library computer. So you should check to see if one or more of the pcs you use has Perl.
You should probably remove the expression "Codemaster" from my nickname, because I found the Perl installation process full of so much stuff I couldn't bear to get into it properly. It semi-worked on the library computer, but when I finally got to play the file, no transcripts came up and I gave up after wasting a good half hour.
Basically I am an artist and writer with a little bit of programming ability. I would rather use my remaining dwindling energy for art and writing. If I were going to make a stab at learning a new language at this late date, it would be Chinese, not Perl.
Here's a question: If using the LoebnerPlayer file is so easy and standard, why hasn't someone else done it and posted the "translation" on the Web?
I decided to go to the trouble of reading the 13-pp. article by Stuart M. Shieber of Harvard which analyzed the first Loebner contest, held in 1991. This sheds light on the attitude of the scholarly/scientific community toward chat robots (Shieber calls them "parlor games"). I also read Loebner's reply to the article. These are both available on Loebner's Web site.
Then I read the 2007 contest rules. It appears that a Pandorabot cannot enter the 2007 Loebner contest, because contestants must send them all bot files, including the bot's engine or interpreter. The Pandorabots engine is closed-source, so that is not possible. Theoretically, I could translate one of my Pbots into an Alicebot, but I have avoided this, because developing a local Alicebot requires much more computer understanding and access than I have.
The only other chatbot contest I know of is the Chatterbox Challenge, but that rules out any ALICE-based bot, and my main character is an ALICE adaptation.
Post by KnyteTrypper on Oct 30, 2006 17:54:55 GMT -5
Actually your eligibility for the Chatterbot Challenge would depend upon how significantly you've rewritten the ALICE AIML. I don't think they preclude other Alicebots from entering, they're just ineligible for certain categories of the competitiion. Then, too, AIML bots are not subject to constraint, just those which use one of the ALICE AIML sets. If your personal AIML rewrite is significant enough that they would not receive any stock ALICE replies, then it is considered to be an AIML bot but not an Alicebot. I think there are already a couple of non-ALICE AIML bots which are regular entrants.
I still want to read the 2006 Loebner Competition Transcripts in plain-text form. I found this "Mailgate" link: mailgate.dada.net/comp/comp.ai.nat-lang/msg30270.html It was somewhat helpful, but the bot and human replies are still all mixed together. Do you know of a posting of the 2006 transcripts in plain text?