Back in the spring, there was talk about forming a group to build a ChatScript bot to compete in the Loebner’s.  While I applauded the sentiment, it was clear to me that one that would pass the qualifiers could not be built in time.

Now things are different.

First, I have won the Loebner’s again. This is with an entirely new chatbot. Back in June, I didn’t have Rosette. All I had was the system that could qualify (based on new script and existing data that comes with ChatScript).

My 2010 Suzette was easily the top qualifier, passing about 50% of the questions. This year’s Rosette was again easily 1st, passing about 16 of the 20 questions, and but for a few generalization bugs in pattern matching due to my haste, would have passed 19 of the 20 questions. Rosette at that time only knew her name and how to pass the qualifiers. For the next few months, she became a chatbot, mostly untested, just ready to enter in October.

So this means the strength is not in “her”, but in ChatScript. ChatScript makes it easy to qualify, and makes it easy to make a strong entrant quickly. Granted I am a master of the language, but to compensate, you have a year and a team!

Obviously I will not be in this team. I have to give you something to topple. But I will be happy to answer questions on how to do things.

bruce wilcox

This has been on the “Back Burner” for me since the idea was originally posted, so you can pretty much bet that I’m in. Anyone else here willing to “take up the gage” with me?

Ok, I am game for this.

I’ve had a mind that there is a rightly configured system that could take the Loebner prize and from my understanding of ChatScript it is the right tech to take this on.

Comment from some other thread about the difference in results between the user groups does not surprise me, the difference in engagement responses with the TeachBot system as opposed to Yhaken or Dorian is self-evident.

Bruce - well done, good effort