I’ve taken a big list of sample questions from two places (here and here) and I’m thinking of using them to develop some pattern matches. I don’t know if I’ll make a bot, I just want to know if I can make something (in PHP) to answer these questions.

I’ve put the questions into a few seperate categories for what “type” of questions they are, to make it easier on me. I was wondering if you think I can get even more specific with the categories, and also if you agree with me putting some of them as “difficult.”

I’ll post them below because it was all too much for this post.

Textbook questions:
What color is the red sea?
WHAT IS THE CAPITAL OF FRANCE
What is a bot?
What is the color of the sky?
What is a cigar made out of?
What is the capital of Spain?
What is the first book of the bible?
What is snow?
What is bigger a dog or an elephant?
What is the location of the Eiffel Tower?
What is YouTube?
What’s the name of the computer who recently competed on Jeopardy?
What month of the year is this?
What is the last month of the year?
Madrid is the capital of what country?
Milk comes from what animal?
Do you know what DVD stands for?
Who is on a dollar bill?
Who is George Bush?
Who is Albert Einstein?
Do u know who Elvis Presley is?
Who wrote the Bible?
Do you know who wrote the Lord of the Rings books?
Do you know how far away the Sun is?
How many days are there in a leap year?
How many stars are there?
How many fingers does a human have?
How many days in a year?
How many sides are there on an octagon?
Can you name two of Earth’s oceans?
Do you think a tomato is a fruit or a vegetable?
How do you eat?
Is snow cold?
When do birds fly south?
When was Y2k?
WHEN DID THE FIRST MAN WALK ON THE MOON
When does sunset occur?
Which is closer the moon or the sun?
Where is Earth?
Where is Moscow located?
Can you tell me the name of a famous actor?
Is the capital of Italy Milan?
Is it possible to prove a negative?
HOW CAN I MAKE A CHATTERBOT
What is a calendar?
How many letters are in the alphabet?

Difficult:
What should you do before crossing a street?
What color do you get when you mix red and yellow?
What is tomorrow’s date?
What language r we speaking?
When should I plant flowers, in the spring or in winter?
Who should I talk to if I have an emotional problem?
Who created the universe?
Name something you would find on a beach.
Name something you would find at the North Pole.
Can you name a famous person who passed away in the last 5 years?
Why are robots better than humans?
Can you give me an example of a living animal?
How is the weather?
Is it winter, spring, summer or fall where you live?
What would I do with a knife?

Profile questions:
How old are you?
Are you male or female?
Are you a bot?
Which do you like better chocolate or vanilla?
WHERE ARE YOU
DO YOU LIKE TALKING TO PEOPLE
CAN YOU TELL JOKES
Where are you anyway?
Do you like Jazz music?
Are you married?
Who is your botmaster?
What is your favorite color?
What is your favorite brand of computer?
What is your favorite drink?
What is your name?
What do you enjoy?
What do you hate?
What country are you from?
What color are your eyes?
What is your zodiac sign?
What is ur religion?
ARE YOU NERVOUS
Who are you?
Where do you call home?
Where were you born?
How’s life?
Who is your best friend?
Where does your botmaster live?
Who are you?
Who is your hero?
If you could be any animal, what would you be?
When were you born or created?
Are you left-handed or right-handed?
Are you homosexual, heterosexual, asexual, or bisexual?
Would you mind if I called you Al?
So Al when is ur birthday?
Do you have a significant other?
My name is Bill. What is your name?
Do you have any brothers or sisters?
What is your favorite food?
Are you a human or a computer?
What’s your favorite color?

Difficult:
WHO IS YOUR FAVORITE STAR WARS CHARACTER
WHAT DO YOU THINK ABOUT GEORGE BUSH
What actress do you think is the prettiest?
What do you like to chat about the most?
What do you want out of life?
What do you think the world will be like in 20 years?
What’s the last dream u remember having?
What do you think is your biggest weakness?
What are you planning to do today?
So what did you get for Christmas?
What kinds of things do you know?
Have you entered any contests?
Wouldn’t you rather be human?
Do you believe in ghosts?
When do you normally go to bed?
Describe your feelings to me.
How does your brain work?
Who is your favorite character in The Wizard of Oz?
How many questions do people normally ask you every day?
Got any plans for the rest of the day?
Do you play any musical instruments?
Have you ever been to Mexico?
Do you believe in Santa Claus?
Are you going on vacation this year?
Approximately how far is the nearest gas station from you?
What time do you go to bed at night?
Do you prefer cats or dogs?
Have you read a good book lately?

Feelings questions:
How is it going?
How are you doing?

Calculations:
What is 1+1?
What is 5 plus ten?
What is 18 + 21 + 3?
What is ten plus fourteen plus eighty one?
What is half of 8?
What number comes before three?
What number comes after twelve?

Difficult:
If I was born on February 23, 1980 how old am I?
Which is larger, a pineapple or a pine tree?
John is taller than Mary and Mary is taller than Sue. Who is shorter, John or Sue?

Memory questions:
What is my name?

Difficult:
My friend Bob likes to play tennis. What game does Bob like to play?
The ball was hit by Bill. What did Bill hit?
Who hit the ball?
What is the name of my friend who likes to play tennis?

Real world questions:
What is the date?
What season is it?
What day of the week is it?
What will be tomorrow’s date?
What time is it?

Verbal manipulation:
What is the opposite of indoors?
What is the opposite of fast?
What is the last letter of the alphabet?
What rhymes with inn?
What word rhymes with cat?
How many is 1,000,000?

Other questions:
CAN I HAVE A PICTURE OF YOU
What is the best nation in the world?
Will you teach me something?
May I tell you a joke?
What were we talking about?
Why do some people think that artificial intelligence is so interesting?
If you were me, would you ask a follow up question?
What do you think it would be like to be human?
Y did the chicken cross the road?
What sound does a dead cat make?
Who will win the 2011 Chatterbox Challenge?

There are also hundreds of questions that most bots should hopefully be able to answer here:
http://www.chatbotbattles.com/

As for writing a chatbot in PHP, this has already been done a few times, unless of course you are asking just for the programming exercise?

The vast majority of those questions are quite familiar to many of us here, long though the list may be. Creating a program (chatbot or otherwise) to answer them is reasonably simple, in and of itself. The “trick” is to create something that can answer those questions accurately, along with the huge array of other similar or related questions with the least amount of code possible. That, my friend, is a rather daunting task, but could be well worth the effort if you’re up to it. Like Steve has mentioned, there are already solutions to this in PHP (imperfect though they are), but it’s a good way to hone and expand your coding skills. Trust me, as lead developer for the Program O project, I’ve had no small experience with just that.

I’m always happy to assist with advice, guidance, or even specific examples, should you be interested, so please feel free to ask questions.

You can find the 1000 questions any mobile assistant should be able to answer here:
http://www.answerdevices.com/index.php?/page/1000_input_mobile_assistant_benchmark.html

Or, you could just create a chatbot and have it chat with another bot for a good long time.  It’s against the rules, but that doesn’t seem to stop others from doing it.  But… I digress.

Regarding your initial post, I’m always suspicious when people are deliberately vague.  No one goes through all that work without a goal in mind. As to your question about types of questions, categories, and difficulty, you’d have to offer a reason for prioritizing them before I could answer.

However, I tend to agree with Dave in that the real trick is in answering “the huge array of other similar or related questions with the least amount of code possible.”

Any bot can be stuffed with a massive amount of information.  But, if you’re interested in something like the Loebner, as you indicated was one of your sources for your question list, then you’ll want your bot (or whatever you’re creating) to anticipate a follow-up question, and to deliver it in something like a conversational tone.  You might also wish to keep in mind that not all input is in the form of a question, frequently, it can be a statement, and a response is expected.

Regarding your list, other than “Are you homosexual, heterosexual, asexual, or bisexual?” you’ve left out the most asked about topics, “sex, cybersex,” and of course, “your mother”.  And then, don’t forget the vast array of wild-card questions, as well as the insults, and all of the name-calling.  Before delving any further into this project, you might wish to gird your loins.

A lot of the “1000 questions any mobile assistant should be able to answer” seems fairly logical…but;

“Why is cuba so poor”

LOL That came out of nowhere.
I suspect that conversation went something like this.

[user][inhales] sssssssssssssppppp

User: Dude….whys cuba so poor?
User: Order me a pizza

“Any bot can be stuffed with a massive amount of information.  But, if you’re interested in something like the Loebner, as you indicated was one of your sources for your question list, then you’ll want your bot (or whatever you’re creating) to anticipate a follow-up question, and to deliver it in something like a conversational tone.  You might also wish to keep in mind that not all input is in the form of a question, frequently, it can be a statement, and a response is expected.”

Actually, the Loebner’s is a schizophrenic contest. You have to qualify by answering questions, which means there is no need to anticipate a followup or deliver a conversational tone.  THEN, if you qualify in the top 4, you have to have a conversation with the judges, which has nothing whatsoever to do with question answering per se.

Bruce Wilcox - Jun 6, 2013:

Actually, the Loebner’s is a schizophrenic contest. You have to qualify by answering questions, which means there is no need to anticipate a followup or deliver a conversational tone.  THEN, if you qualify in the top 4, you have to have a conversation with the judges, which has nothing whatsoever to do with question answering per se.

Having won the the contest, you’d know more about it than I would, Bruce, so I’ll defer to your wisdom.  To that, I’d add an additional level confusion in that (according to an email reply from Dr. Loebner) “there are no rules.”

It is correct that there are no rules “during the Judging”, but not during the qualification round.

And recently since I fooled a judge with Suzette, they are now careful to ask questions a chatbot will not likely get right like “if I stab you with a towel would it hurt”, but that’s just to insure they know which are chatbots and which are not. After that, its just a conversation, to determine which chatbot is “most human” even though obviously not human.

“However, I tend to agree with Dave in that the real trick is in answering “the huge array of other similar or related questions with the least amount of code possible.””

I agree with Thunder Walk who agrees with Dave, although I would settle for a “reasonable amount of reusable code” instead of the “least amount of code possible”.

If we address a subset of the questions above for a specific example:
What is your favorite color?
What is your favorite brand of computer?
What is your favorite drink?
WHO IS YOUR FAVORITE STAR WARS CHARACTER

Then rather than trying to come up with every possible way someone could ask you about what your favorite thing is I would be interested in the strategies others have used to respond accurately using a “reasonable amount of reusable code” (reasonable = almost anything besides hardcoding a fixed response to each possible question using AIML.)

I would define “responding accurately” as being able to respond uniquely to each question.  (and not just testing for the word “favorite” and then replying with the template “I don’t have one. What is your favorite?” or other such general catch-all responses.)

I would think one strategy might be to store each “fact” as a aiml category or as a triple in a datastore (me have_favorite_food cake) and then have a method to look up the fact and respond using a variety of sentence patterns.  The possible ways the questions can be asked can be mapped to the same method so that questions such as “Do you have a favorite x?” would be processed the same as “What is your favorite X?” as would “You must have a favorite team.”  I would expect the question “Does your dog have a favorite food?” to be answered differently.

What strategies have people used in a chatbot to answer similar or related questions in the case of “what is your favorite x” type questions above.  Is anyone willing to share their strategies for this case?  And is there a strategy that allows for the answering of specific questions but also allows for use in general conversation such as in the schizophrenic Loebner contest requirements as Bruce states above?

In ChatScript I have a topic which recognizes the various ways one might ask about a favorite x or a favorite x y, and I have data tables that create facts that can be queried for them.  So a fact table for beverages might have an entry for

_ tea “I love jasmine tea”

herbal tea “I love chamomile tea”

_ ~wine   “I love Shiraz”

and the topic detects the underlying one or two word request and retrieve the appropriate data from this table. If the question was what is your favorite “green tea”, since there is no entry for green tea, it would settle for _ tea.

This is equally available in qualifier and conversational modes.

I have tables of “favorites”, of “hates”,  of “can you” or “do you”.

Bruce Wilcox - Jun 6, 2013:

You have to qualify by answering questions, which means there is no need to anticipate a followup or deliver a conversational tone.

This is mostly true but recently they have asked questions such as “What is my name?” and then asked a follow up question like “How many letters in my name?”
Another example is saying a statement like “Fred kicked the football.” and then asking “What did Fred kick?”

The method of questioning in the final 4 explains the radical difference in results when the Junior Loebner Prize took place in 2011. The school children were talking to the bots like friends, asking them what their favorite TV show was and if they liked the latest boyband, whereas the adult judges were blasting them with “how many syllables in balloon?” and “Which is bigger a whale or China?” from the outset without even saying hello.

Organizing and conducting as contest must be a nightmare, but not impossible.  I think the Loebner would benefit from a longer time limit for the testing period, allowing for something like a conversation to form.  In 1991, judges moved to a new terminal after “about” 18 minutes.

I also imagine that the selection of interrogators, or judges, could be improved upon.  In 1991, 10 judges were selected from a pool of those who answered an add placed in a Boston paper for people with a lack of knowledge of computers.  That method sounds odd today, given the proliferation of home computers, but it seems there must be a way of gathering a panel of judges somewhere in between hard-core computer people and the school janitor with no interest in computers.  It might also help if the interrogators knew something about conversation… or interviewing techniques.

I’d also think there would be some benefit to having boundaries, or at least, topics.  In 1992 they were Health, Hockey, Relations With Your Real or Imaginary Wife or Husband, Men vs. Women, Dream Analysis, and the Boston Red Sox.  The transcripts make for some interesting reading.  I don’t think we’re ready yet for a bot to converse about just anything that comes to mind.

“.  I think the Loebner would benefit from a longer time limit for the testing period”

The ChatbotBattles contest of Steve’s had a 15 minute conversation format as part of it, which was sufficient for “having a conversation”.

“What is my name?” and then asked a follow up question like “How many letters in my name?”
Another example is saying a statement like “Fred kicked the football.” and then asking “What did Fred kick?”

Yes, of course it requires you learn data from during the questioning to answer more questions.  I wasn’t thinking of that. I was thinking of the usual sort of “followup” where the followup question depends on the original answer. Eg wherein I might say “do you like flowers” and you answer “I love roses” and then I follow up with “which is your most favorite color”.

I suppose the time limit would depend on the questions.  If there was a topic, or it was a legitimate conversation rather attempts to trick the bot, or to simply test it to see how many obscure things it had been fed (like say, a list of locations for every area code) then it might be shorter.

Given the aim of the contest, I’ve often wondered why the chats weren’t more like what would occur between two people sitting in a waiting room of some sort before a scheduled appointment, or strangers seated on a park bench getting to know each other.

If a stranger approached me and asked, “What would I use a hammer for?” or “John is older than Mary, and Mary is older than Sarah. Which of them is the oldest?” I think I’d run in the other direction.