Bug Stomp

The (key1) is not being inserted here:

human: Well, I'm underage. And I'm from England.
Love Agent: Oh, is nice I hear in the spring. But where do you live now?

The keyphrase being triggered here is:
I am from * , I grew up in *

The response is:
Oh, (key1) is nice I hear in the spring. But where do you live now?

You might try changing the keyphrase to "I am from (*)" or "I am from" (in which case the response would refer to (postkey) rather than (Key1).) I suggested the former first because of the current discussion about (postkey) not being reliable in some cases.

That's actually correct.

There's really no point to having a * at the end of any Keyphrase. That's a soft wildcard, and will always be blank when used at the end. (postkey) was made to contain the part of a sentence that follows the Keyphrase.

To understand the soft wildcard, it is built to match as little as possible. This is known as being "ungreedy", and it was very specifically built that way. Take a look at this example:

Keyphrase: i like * the
Sentence: i like the way the cat feels.

Since the soft wildcard is ungreedy, the Keyphrase will match "i like the" instead of matching the SECOND "the" in the sentence, which would be "i like the way the" and would lead to all sorts of unforseen results.

Does that make sense?

Perfect sense here. Actually, I think I had problems with that when I first started using wildcards and ended up trashing it because it never worked.
Thanks for explaining! Now I know what I did wrong...

This one bot, Miki's Vanessa, is getting its name a little messed up in the transcript. It's a fairly new bot so I don't know whether this really affects keyphrase rank and things like that, but the Vanessa part shows up every time as the first word the bot says in each exchange.

That makes sense. Thanks Prof! And now for another mystery... I think ranks are not being calculated correctly:

You: what did you study?
Bot: Sure I did!

In this case, Love Agent is responding with the keyphrase for "did (not|) you" ranked -5, while the phrase "what * you study", ranked 50, was left untriggered.

What about an option to make * greedy?

like... "*!" or % or whatever regexp uses

Importing seeks don't work if it's like:
+yes[0]

it needs to be +yes [0] or + yes [0]

could we maybe have it as acceptable, or would that break things?

ERROR: All Seeks must be intended and follow a Response.
Line: "+no [0] "

I think prof meant indented

Well, you wouldn't want an unintentional Seek, either.

I'm having a problem getting Duo to recognize the following keyphrase:

Would you like to (guess|know) how old I am, Can you guess how old I am [25,0]

Whenever I test him, the following occurs:

Me: Would you like to know how old I am?
Bot: I didn't realize I was that obvious.

In here he's actually replying to the keyphrase "You like [2,2]", which is ranked way lower than the above.


But when I ask:

Me: can you guess how old I am?
Bot: Not a good idea. If I guess too young you'll think I'm calling you immature. If I guess too old, you'll think I mean you're boring.

- then he actually replies to the correct keyphrase. I'm not sure what's causing the problem, though. It seems like the gap between the ranking of the keyphrases is large enough to point him to the right one in both instances.

/puzzled/