[alicebot-developer] Re: Program verification standards?

[Josip Almasi]

David Sienkiewicz wrote:
> 
> Josip: what I was hoping for was a concise set of points with which I 
> could benchmark the programs various functions. I thought it might take 
> too long to just allow a random conversation although I could see some 
> merit in doing so. I also didn't want to step through each and every 
> template and pattern in the triple-A set if I didn't have to.

Yep, I guessed that much - you're looking for some kind of 'unit tests', 
right?
Well, tough luck... I don't even understand how such a thing could be 
done with AIML.
Eventually some kind of brute force, i.e. random strings or random word 
combinations based on some grammar...

> What would be of interest is some sort of a modified ALICE program that 
> has the sole purpose of evaluating and rating other ALICE programs, 
> perhaps not only for content but maybe response time as well.

Again, I don't see how you can evaluate the content, in general.
How can you evaluate english, russian, turkish bots with one program?
But this _is_ an interesting idea... if the Foundation had such a 
'conformance test', it would make interpreter implementation much easier.

> To answer your question, not Russian (although close), Polish. 
> ...unconfirmed distant relation. The folks that knew for sure died 
> before passing on the information in a way that made sense. We do have 
> one tantalizing first edition book in the family.

Well then, FYI: I've met writer's direct descedant. The familly ran away 
from russians to belgrade.

Regards...