Members don't see the ad below. Register now!

# How does akinator work?

 2 Hey, i've been thinking: how does the akinator work? (en.akinator.com) I tought about some answers and i even did a simulation (took me four days), but i wish to know what do you people think about it. (cause my simulation doesnt work very well, especially the part about the question select, where the intelligent agent has to select a question to ask to the client) Sorry by the bad english, but i think my text is understandable asked 15 Oct '11, 17:03 Macmod 31●1●1●3

 3 The best question to ask would be the one that brings it closest to the goal. The goal would be to have only a single remaining possible candidate, and so a good heuristic would be the number of possible remaining candidates. Therefore, your best bet for approaching the goal directly would be to reduce the number of remaining candidates as much as possible with each question. Since the problem is stochastic (he doesn't know whether you will answer yes or no), the way to ensure the remaining group is as low as possible is to split the group down the middle as closely as possible. The more uneven the split, the more chance that your answer will be in the 'larger' half, and the greater the list of remaining candidates, on average. (Although even ignoring the fact that your answer is probably more likely to be in the 'larger' half, this method of splitting ensures reliability.) Thus if half the people in his database were men, and half were women, then that would be the best candidate for the first question. By this reckoning, he would be accurately able to guess 2^20 people, under optimal conditions. answered 16 Oct '11, 05:52 Fishy ♦ 2.8k●7●25 1 An excellent description: partitioning the knowledge base in such a way it can minimize the number of questions. I expect that a system like this, which has been online and used by a wide variety of people, also has statistics on the most common choices. Partitioning the space so common choices are focused on for elimination or confirmation would accelerate the rate of convergence. (17 Oct '11, 17:43) Godeke ♦
Members don't see the ad. Register now!
 1 The Akinator is a smart phone and web based game that can determine which character a player is thinking of by asking him/her a series of twenty, or less, questions. You can think of it as the Binary Search Algorithm. In each iteration, we ask a question, which should eliminate roughly half of the possible word choices. If there are total of N words, then we can expect to get an answer after log2(N) questions. With 20 question, we should optimally be able to find a word among 2^20 = 1 million words. One easy way to eliminate outliers (wrong answers) would be to probably use something like RANSAC. This would mean, instead of takinging into account all questions which hae been answered, you randomly pick a smaller subset, which is enough to give you a single answer. Now you repeat that a few times with different random subset of questions, till you see that most of the time, you are getting the same result. you then know you have the right answer. Ofcourse this is just one way of many ways of solving this problem. answered 16 Oct '11, 06:54 aarvy_ravi 76●1 Wikipedia has an article about RANSAC: http://en.wikipedia.org/wiki/RANSAC (17 Oct '11, 17:27) EllenL
 0 When i was learning about AI before, we implemented something called Expert System, it is used to make medical system or for techinal support. It is something that learns from new information and it is usually used as a classificator. I think Akinator uses that. answered 16 Oct '11, 06:32 juansev 14●3
 0 When i was learning about AI before, we implemented something called Expert System, it is used to make medical system or for techinal support. It is something that learns from new information and it is usually used as a classificator. I think Akinator uses that. answered 16 Oct '11, 06:33 juansev 14●3
 0 It's simply an Expert System. The system compares your input with a database of facts, and infer based on predefined rules. You can easily implement something similar to akinator with CLIPS. It's relatively easy to implement the system and the rules, the problem resides in the database of facts, because for your system to be accurate and efficient you need a huge database of facts. answered 16 Oct '11, 09:28 lrq3000 2.0k●3●17
 toggle preview community wiki