AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

history of Suzette and issues in chatlanguage design
 
 
  [ # 16 ]
Dave Morton - Feb 9, 2012:

If anyone is interested, I’m more than happy to post the PHP code used to create the transcript file, so that you might use it to transcribe other log files.

Good idea! Post it in the Loebner forum. It will inform Hugh as well, so he might even give some suggestions..

 

 
  [ # 17 ]

You wish.  Humans can reply w/o waiting for enter key from computer (though rare). and my bot will respond w/o cr if human waits long eenough between characters. Happens often enough… human types a few words, thinks for a moment or is distracted, my bots reply, then human enters more with enter key, that’s a new burst back, unrelated to what my bot just sent

 

 
  [ # 18 ]

@Erwin:

I’ll have to revise the code a bit before I do so. At the end of the Loebner Competition “official Rules” page, I saw a table of “Special Characters” that will help me immensely in processing the log entries. All I had to work with originally was the one raw log file, so there were “pieces” missing. the new algorithm will be much more streamlined.

@Bruce:

I was noticing that while I was reviewing the logs, using the “Loebner Player”, which I hadn’t found till the transcript processor was well under way. But that’s alright. When comparing the output that my script created with the output from the Loebner Player, the differences were very minimal, and there were no discrepancies based on timing, so I’m happy with the outcome.

 

 
  [ # 19 ]

While refining the script and adding a feature (loading different log files would prove helpful, I think) I noticed a discrepancy between the 2010 log files and the 2011 versions, where the older versions all listed the chatbot entities as “program”, whereas at least a couple of the 2011 logs listed the chatbots by name. I think I can pretty well work around this, but “changing horses in the middle of the stream” adds another layer of complexity to the script that will add to it’s completion time. Of course, if it were easy, it wouldn’t be as much fun, so I won’t complain… Much. cheese

 

 
  [ # 20 ]

I should probably point out here that there is already a better solution to the problem that I’m tackling, in the form of the Loebner Log Player. It’s a stand-alone Windows program that allows you to see the log files “played out” in “real time”, and is far superior to the solution that I’ve got, in nearly all respects. I’ll still work on this little project, but I can’t see where I’ll be able to improve on that program, except for the tiny little aspect that mine will be available as a web page when I’m done. just thought you may wish to know. smile

 

 
  [ # 21 ]

Dave, I look forward to your project. The log player is tedious—I’m not so much interested in the real-time aspect as in reading the final logs. I realize the typing component is important for those wishing to participate in the Loebner competition, but for those looking for just the overall flow of the conversations, a transcript is much more straightforward.

 

 
  [ # 22 ]

I have to agree, CR, that the player is not my preferred method of viewing the transcripts; otherwise, I’d not still be working on this. I’ve uploaded what I currently have to my website, and you can view it at http://www.geekcavecreations.com/LoebnerLogs/hrTranscript.php - The 2010 logs all seem to work fine, but the 2011 logs still generate some screen garbage, so I have to figure out what’s happening there.

 

 
  [ # 23 ]

@Dave: OK, I’m patient grin

 

 
  [ # 24 ]

Good work Dave!
The 2010 Carpenter and Medeksza don’t have the responses highlighted correctly.

 

 
  [ # 25 ]

Oh, GREAT! You broke it, Merlin! raspberry

lol, kidding. {sigh!} it works fine, locally. Perhaps the script got corrupted somehow, during upload, or there’s some strange “windows vs. Linux” problem, somewhere. I’ll look into it.

 

 
  [ # 26 ]

I can understand in some respects why the single character requirement output, but on the other hand it does not qualify the bots abilities for realistic AI. It seems more like a programming challege to work around the requirement than a test of your bot’s logic capibiliies. The real human has it easy, it’s the poor bot that has to work around the chat language obstacle course! wink

 

 
  [ # 27 ]

That is one of the reasons I have not entered so far.
It does not mimic any typical communication system and the development ends up being a 1 off throw-away. Hard to justify when there are other things that could make your bot better.

 

 
  [ # 28 ]

True Merlin.. development time for your bot is gold when you also have a daytime job. By studying Suzette’s raw chat files, I see how I could adapt Marie’s output to the proper file specification, but on the other hand, I am not so sure that it justifies my time.

 

 
  [ # 29 ]

You both may wish to have a chat with Dr. Rich Wallace from Pandorabots. I can’t guarantee that he’ll be able to help, but he may have some useful input regarding the LPP, and it’s possible he may be able to point you toward a solution that will save some time. It certainly couldn’t hurt, I’m thinking. smile

 

 
  [ # 30 ]

I set up a Turing test for my Mitsuku bot at:

http://square-bear.co.uk/mitsuku/turing/

and most of the people who think it is human are of that opinion due to the spelling mistakes, variable typing speeds or backspacing I intentionally put in. I would say to make a convincing Turing test bot, the single character input is a very important element.

 

 < 1 2 3 > 
2 of 3
 
  login or register to react