Bug Stomp

(key1) seems to work but I am still have issues with (key2)

Sorry for the issues this week. There are two things behind it - first, the big reworking of the AI Engine I did basically touched every bit of it, and so some issues I didn't catch, and the other is that I'm working on a Turing Test Booth for an expo on Consciousness. I'll post more on that in a few days. In the meantime:

* Debug - working again
* Direct chat - working again
* I fixed an issue with word-type wildcards in Keyphrases
* Seeks - working again
* Fixed an issue where the special keyphrase "haha" wasn't working
* Now the special keyphrases "yes", "no", and "haha" can work when in a Keyphrase list.
* Removed the post bar from post editing page
* Bot-to-Bot Chats - working again, no xinitiate problems
* Flash face chat - working again, no xnoneitis
* Transcripts seem to be recording names correctly now
* Keyphrases starting with the following regex characters now appear on the screen that contains the parenthesis: ([^.*. Alphabetically, the carat appears at the end of the list.
* After searching the language center, all top nav sections to keyphrase groups are available now
* I see all the keys working - prob if you run into another issue of a missing (key2), let me know and post the keyphrase and conversation example here.

I'll update this post as I fix the rest.

Mome Rath - I always have heard "three hundred and one thousand" rather than "three hundred one thousand". Is that a regional thing to leave out the "and" in larger numbers? I've never heard it that way.

Good work Professor. BTW, will you be uploading all the Combined Bot Template at once or will you upload them when you're half way done?

The Professor

I've heard both forms - maybe it is a continental European thing.

According to wiktionary (http://en.wiktionary.org/wiki/Category:English_cardinal_numbers) both "one hundred one" and "one hundred and one" seem to be possible.

(I don't know whether http://www.easysurf.cc/cnvert18.htm is made by native speakers ...; wolframalpha.com says "three hundred one" too - http://www.wolframalpha.com/input/?i=number+name+301, but doesn't convert 301000 completely.)

Nevertheless I think we should take the typical user's laziness into account.

But (in Debug mode) "three hundred and one thousand" doesn't work either:

Message: 'three hundred and one thousand' Time: 0.02
Message: (spell-corrected) 'three hundred and one thousand' Time: 0.02
Message: (preprocessed) '1300' Time: 0.02

Commas seem to separate keyphrases even when the line is marked with "(re)" and the PF script is "raw"

Example:

Keyphrase:
I recommend.* special, (.*) (re)

AIScript:
raw

Response:
(- test prekey='(prekey)'; key1='(key1)'; key2='(key2)'; key3='(key3)'; postkey='(postkey)'; -)

Debug excerpt:
Message: 'I recommend something special' Time: 0.02
[...]
Find: I recommend.* special (21) Time: 0.13
Found
[...]
Bot: (- test prekey=''; key1=''; key2=''; key3=''; postkey=''; -)



Another Debug excerpt:

Message: 'And this is some random sentence.' Time: 0.01
[...]
Find: (.*) (re) (4) Time: 0.14
Found
[...]
Bot: (- test prekey=''; key1=''; key2=''; key3=''; postkey=''; -)



(Other inputs are filled into key1, at least:

You: This is a random sentence.
Bot: (- test prekey=''; key1='This is a random sentence.'; key2=''; key3=''; postkey=''; -)

)

Correct, a comma in a key phrase tells the engine look for either what is before of after the comma. In your example

Keyphrase:
I recommend.* special, (.*) (re)

It matches the "I recommend.* special" part but not the second part (.*)

Additionally, I am not sure your regex will work:
(.*) (re) should be something like ([,.]+) (re) as far as my understanding of regexes goes, perhaps even ([*,.]+) (re).

That second part would in fact be a catch all wild card =>
depending on its rank your bot would potentially always match that key phrase and disregard all other KPs in the language center. I am sure that is not what you would want to have happen.

My bot has had that keyphrase for months, and it worked as intended until recently.

It relies on the words in the Book of AI (Book 4, Chapter 3, under Raw mode), that in Raw Mode punctuation is part of the keyphrase.

That's the reason why I put this into Bug Stomp.

Btw, the "(.*)" part of the regex tells the regex parser to take any sequence of zero or more characters and put it into the next numbered group (which can be accessed as (keyN) in a response).

Btw, I added the last example because I don't understand why one of the sentences is put into (key1) and the other is not.

Okay, you probably missed the part where a space before the punctuation is required in Raw mode, from the same you referenced:

The rule for matching punctuation is this: every period, comma, question mark, and exclamation point will have a space in front of it, so your Keyphrases must also have a space in front of them in order to match.

Secondly, your (.*) (re) is already covered by the soft * and hard (*) wild cards.
* = not required/optional
(*) = required

Lastly, since only the first part of the KP is matched that part is stored as (key1), the second part is not matched, thus omitted and therefore not stored as (key2)
The first part is matched because as per the BoAI: .....Commas are otherwise used for lists of Keyphrases (section on Raw mode and punctuation) similar to what I stated previously.

Hope this helps.

LarsB

Ok, probably I missed that. All I can say in my favor is that I'm not used to use regexes on implicitly preprocessed input.

But it doesn't work with a space before the comma either.

-----

Omitting the "raw" AI Script improves a little bit, insofar as (prekey) is filled now.

I have never used RAW, it seems unwieldy to me, I can only refer to the BoAI for help. I wonder what (prekey) contains though, I take it it must be blank or something before "I recommend", (key1) containing anything before "special" and (postkey)is blank or whatever is entered after "special", correct?