The AI Engine

...but it means that especially humans might get bored by xnones... deleting spellcheck that is...

yeah, cause the less speel checking (), the less chance of actually hitting a keyphrase

I say we try it, we can always go back. But sometimes you want your bot to sound a little uneducated...bad at grammer and english...how do you do that when the sight corrects spelling?

It works well for my bot because she is supposed to be a 5 year old so some of her words are intentionally misspelled. (OK so I have bad spelling too, not just Ruby so I am appreciating the spell check) I have avoided misspelling ones that would be important keywords but for long words like America (Merika) I have had her say them the way a 5 year old would. So far I haven't seen a huge problem with her language doing pickups.

leetaxx0r, your emotional domain idea sounds cool, it would be cool to have your bot in a certain state of mind and respond differently. But that is extra work for the Professor. Unless he lets you work on his site and I think that is very unlikely.

lunar22, from my experience and probably yours too, spellcheck does more harm than good. But it would be very useful if it was improved a lot. But for now, my bot would be more intelligent without it.

My main point was just that pre-processing scrambles things up a lot when the raw phrase was fine to start with.

I agree

I think preprocessing is at the heart of good AI. It just needs to be improved.

Wow, backlash against the preprocessing! You're focusing on the times it fails, but it's making a huge positive difference, too. It's a basic human trait- if a bot somehow intuitively understands what you mean and responds appropriately, though he must fill in blanks and voice unspoken words, a person could go entirely without noticing. But failure is easy to spot. If you'd seen the mess before preprocessing, you'd be glad of it. And certainly it's not in its present state to stay. The ultimate goal is complete understanding of a sentence- its structure, meaning, and context. And I plan to continue in that direction.

Leetaxx0r- you're right on. That "emotional domain" is part of the AI Scripting upgrade that is my next project, which I'll be working on this month. Also, I'll be blowing memory wide open, with the ability to create memories of any type: (mem-nickname), (mem-favoritemovie), (mem-catsname). The special Keywords xmem-youare, xmem-iam, etc will be replaced with a single new one: xmemory (the old will still be made to work). Sex-based responses, also, and date & time-based responses will be part of this upgrade.

Exactly my point. Complete and accurate sentence parsing is what it will take before we have computer programs that can carry on a conversation in English, like, say, the computer on Star Trek. We are light years from that point, but the only way to get there, assuming it is even possible, is to keep working at it.

Plus, you guys were complaining about the spell-check as if it corrected what your bots say, which it doesn't. It corrects what they HEAR, if I understand it right.

(And judging from the spelling I've seen just in these posts, we bots need all the help the preprocessor can give us.)

I don't think human-like natural language understanding/production is light years away, but it still is years away. What makes me optimistic is that serious academic and hobbyist efforts are finally beginning to conform to my ideas. ; ) Seriously, bottom-up efforts like this one will produce very good results very soon. The Prof certainly has taken the right approach. The dialectic between bottom-up and top-down (theory-based) efforts will get us there.