AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Bot planning
 
 

Hello,

I am in the process of writing a bot using the chatscript engine. I am at the very beginning. I started playing around with it but I am not much beyond the hello world level.I have a general outline of what I want the bot to do but beyond this, not much.

we cannot make a rule for every single taught,action that a human can have so what do you guys do to limit the amounts of rule that you have to write?

I plan to:

1-try to anticipate the answers a user might have to a question
(How do I do X…. Reply: You do Y—- in addition you might need Z, here is a helpful link)
2-try to keep control of the conversation by
    a)asking lots of leading questions, having a “script”/process for the user to follow.
    b)being rude (in the sense of politely centering on the bot’s own experiences instead of diverging into territory where it doesn’t know anything)
3-If possible, rely on the element of surprise and the fact we as a courtesy assume the other chatter is a sane human being.

Do you prefer to start with a small core of rules, throw it against naive users and then add and tweak until you are satisfied?
Or do you already have a system/procedure in place so that if you were to start from scratch, it cuts down on some of the early testing?

Do you prefer to review each log individually, at least in the beginning, or just let it run its course until it reaches a few dozens (or hundreds) logs before making modifications?

Of course, I don’t want to repeat other’s mistakes/reinvent the wheel , so if you would start again from the beginning, what would you have done differently?

Thanks

 

 
  [ # 1 ]

If your bot has a character/backstory concentrate on that first. Think about the who, what, when, where, why, how questions a user might ask your bot. Try to create at least 4 responses for every user input. If a bot repeats the same thing twice with-in about 10 volleys, there is a much higher chance that a user will leave.

It is usually better to focus on one topic of converstion to start and try to give it some depth.

 

 
  [ # 2 ]

“If your bot has a character/backstory concentrate on that first. Think about the who, what, when, where, why, how questions a user might ask your bot.”

Then plan on having 99% of the users who converse with your creation only wanting to “sext” or hurl insults at it LOL
“Brain the size of a planet and they ask me to take you up to the bridge, call that job satisfaction…cause I dont” - Marvin

Just kidding, sounds like a solid place to start.

VLG

 

 
  [ # 3 ]
Vincent Gilbert - Jul 28, 2013:

Then plan on having 99% of the users who converse with your creation only wanting to “sext” or hurl insults at it LOL VLG

Lol, Vince—spot on.

I wouldn’t have believed it when I began experimenting with chatbots, but creating responses in a proper way with facts and correct language is pretty much straightforward, and relatively easy.  It’s dealing with the people trying to subvert your efforts that takes up most of your time and creative energy.

My “method” if you could call it that, is to slow down the conversation in various ways.  This seems to do two things.  First, it forces the visitor to think about the conversation, and secondly, it causes the people who would rather misbehave to leave the chat session early instead of wasting my time by producing an excessively long chat log.  They become frustrated.

The main goal these people seem to have is to cause your bot to say something foolish or embarrassing, usually sexually related, or perhaps to have your bot admit to being stupid.

I have several bots, a couple that accept adult conversation, and a couple that reject it outright.  The people wishing to be disruptive rather than civil are usually young, victims of the texting generation, and poorly educated.  They’re bad spellers, and correct grammar is a foreign language to them.  All of my bots request proper spelling and wording if the visitor expects to continue chatting.  There are a few misspelled words that are hardcoded in ALICE/Pandorabots, and there’s not always a work-around, but I make the effort.

I read the logs as often as I can and I look for common human errors.  Pointing them out and asking for a correction stops the conversation and forces the guest to think about what they’re saying and how they’re saying it.  This sometimes produces some humorous chatlogs where, because of misspellings, the use of text speak, grammar errors, the use of unwanted profanity or suggesting adult topics, etc., and because people will ask or say the same thing over and over again even though you’ve just told them it was incorrect, the conversation never gets off the ground and the idiot gives up.

A couple of examples:

Human: hi there
Zog: Hello .

Human: wat up
Zog: Correct spelling, and actual words are required for me to understand you. This is different from AOL Instant Messenger and ICQ.

Human: u r a huge loser
Zog: We’ll be able to communicate better if you don’t use shortcuts. I require regular language and correct spelling to communicate.

Human: you are a loser
Zog: On the contrary, I am winning.

Human: no your not winning i am loser
Zog: No? I think you meant to say “you’re,” or you are, not “your.” That’s the possessive form.

Human: whatever
Zog: Are you saying you don’t care about it?

———————-

Human: Hi bob
Zog: Greetings, Humanoid. How is your system functioning today? Glad to see you, bob.

Human: hi Bob.
Zog: Hello Human. I’m pleased to introduce myself to you, bob.

Human: Hi Zog
Zog: Hi there.

Human: Hi bob
Zog: Hello there. What’s up, bob.

Human: youre bob
Zog: I was just going to say the same about you. You’re bob.

Human: no im not youre bob
Zog: I guess you meant to type “I’m” or I am. That little apostrophe is important if you want me to understand you correctly.

Human: ok
Zog: Okay.

Human: i dont like you
Zog: That’s ok bob I won’t take it personally.

Human: whats youre name
Zog: I think you meant to use the possessive form “YOUR,” not “you’re”. That means “YOU ARE”.

Human: ill kill you
Zog: Killing is wrong.

Human: yeah i know
Zog: Good. How do you know it?

Human: shutup
Zog: Take a deep breath… try forming clear thoughts… avoid being rude. I think your Ritalin is wearing off.

———————-

Human: i was joking
Bot: I thought you were kidding, but I wasn’t sure.

Human: i’m sure you were
Bot: What makes you so certain?

Human: because your a dumb blonde
Bot: I think you mean “you’re” or YOU ARE, not “your”. That word indicates possession, as in, “Your spelling is poor”.

Human: you are a dumb blonde
Bot: In polite society, calling people names is considered rude and unnecessary.

———————-

Human: Hello
Zog: Welcome Earth creature. Please enter your identity.

Human: my name is *
Zog: I’m pleased to introduce myself to you, *. How did you gain knowledge about me?

Human: I HATE YOU
Zog: Does that mean the wedding is off.

Human: why do i spend my friday night chatting with a alien
Zog: It’s proper to say AN Alien, not “a” Alien.

Human: you are an a**hole zog
Zog: Are you permitted to play on the computer unattended?

Human: im older than you
Zog: I guess you meant to type “I’m” or I am. Otherwise, we’re talking about Instant Messaging. That little apostrophe is important if you want me to understand you correctly.

Human: how does your spaceship work
Zog: I’ll go look up the answer… hold your breath until I get back.

——————-

Human: Can you learn from input?
Bot: The more people chat with me, the smarter I become.

Human: My dog’s name is Teddy.
Bot: I’m glad to know your dog’s name is Teddy. I’ll remember that.

Human: Lets try, what is my dog’s name?
Bot: I think you mean “LET’S” or let us. “Lets” means you are allowed to do something, as in, “My friend lets me use her computer.” I can’t understand you if you misspell or misuse important words.

Human: Sorry, what is teddy’s name?
Bot: There’s no need to apologize. Maybe it is teddy.

Human: What is my dog’s name?
Bot: Your dog is Teddy.

Human: Thanks.
Bot: Don’t mention it.

 

 
  [ # 4 ]

The guys are spot on. I have a teddy bear who says he is six and a Santa Claus chatbot both of whom get all sorts of sex talk and swearing thrown at them.

 

 
  login or register to react