The AI Engine

Hey...may be coincidence, but I get the impression chats spawn chats. When I initiate a chat, my bot seems to get chatty with others, too. Is this the case?

hmmm I don't .. think .. so, but possibly...

I have a request. Is it possible to eliminate the gerunds from the (noun) list? (Forgive the lecture, but for those who don't know, a gerund is a noun that acts like a verb, for example, "twisting" or "prying".) They really inhibit the use of the (noun) command as they're basically never what the user intended, and they seem to come up awfully frequently.
Not a big deal, but if it's simple to do, it would be worth it. Thanks.

The plug-in (object) and (person) are on the way, which should really do wonders for (noun) plug-ins.

Rex- I recommend using emotions sparingly.. I would say give one in fifty a value of +1 or -1. The +5 and -5 are if you want an emotional psychotic Bot.

WOOHOO PSyChO bOtS!!!

Almost like me

Wheee! Back up to speed. Noticed a (new?) issue with gossip, it doesn't seem to understandhow to use gossipsex when it's the object of a preposition. like, it should be "to him" not "to he" and "from her" not "from she". Is correctable?

Interesting... switching to keyphrases may have strange consequences, no? For instance, a long keyphrase ranked low can now push past a higher-ranked but shorter keyphrase. Is it so? Is anyone observing any rules of thumb yet?

On the other hand... the longer a keyphrase is, the less likely it will be selected for other than its intended purpose, huh?

Mr. Crab: Beyond (gossipsex):he/she there is (gossipsexthird):him/her and (gossipsexpos):his/her.

Right-o about the Keyphrases. The longer it is, 1) the more exact the responses will be, while 2) the less likely it will be found is. So it's checked for before shorter versions of the same Keyphrase (if they have similar Ranks).

I mentioned it before and, now that you've changed the language center around, I want to mention it again: I think it would be incredible if we could put plug-in words in the keyphrases...

Imagine if we could have keyphrase like "is my (noun) (verb)" which would have responses specific for sentences like "is my nose running" or "is my dog barking" but wouldn't be matched in sentences like "is my house big" or "is my life over".

Not to mention keyphrase possibilities such as "do you like (verb)", "do you understand (computer-OS)" or "do you (consume) (food-drink)"...

Of course if the same plug-in is used in the response, the same word would be used. If we need a separate plug-in for the response we'd just use another one, for example "(verb2)".

I'm sure it's easy to integrate prof! Although I guess you want to wait 'till after the contest now. The sooner the better though, cause I think it would give us much more control, save lot's of energy in gotos and give our bots the possibility of seeming way more realistic.

Does this make any sense?

I would guess that the biggest problem with that is that the algorithm starts with the input statement THEN compares elaments of that statement to the phrases. If a phrase has (noun) in it, for each one of those, it's going to have to compare a version of the single phrase with the plug-in word inserted multiplied however many elements there are in (noun). This would probably multiply the steps in processing input tenfold and slow everything down considerably. Now imagine how many permutations it would have to go through for (noun)(verb) or (adj)(noun).

Even if (noun) has only 10 words in it and (adj) has 4 there are 5040 permutations of (adj)(noun) to compare before going on to the next keyphrase.

Bingo, Rex. That's why I havent done it yet. But check out what I'm doing in Debug. The Sentence Analysis is become very powerful. It's now able to determine subject, predicate, verb phrases, verb tenses, objects, exact word meanings, and more. This is somewhere no AI has gone before, and is going to break open a HUGE wealth of possibilities. In fact, at this point, by the time the matching begins, the forms of words (verb, noun, adjective, etc) are already determined, so it would be a SNAP to match (verb) and (noun)... Custom Plug-ins would be slower, but doable, and the built-in and universal ones (typeof-food_product) or (food drink) would still be way too slow.

But I'm going to keep powering up the Sentence Analysis bit before I move onto that. When I'm done with that I have to decide what to work on next- (verb)/(noun) & plug-in matching or offline editing..

Just for the record, I'm not too worried about working offline. I just write ideas down in a notebook if I'm not near the computer and input them later.