Brain Teasers
Brain Teasers Trivia Mentalrobics Games Community
Personal Links
Your Favorites
Puzzle
Hang Glider
Sudoku
Beanstalk
Pirate's Booty
Word Search
Code Breaker
Math Path
Disco
Tetris
Battleship
Sliding Tile
Twiddle
Strategy
Cookie World
Food Fight
Chess
Checkers
Werewolf
Who's the Boss?
Entertaining
Pac-Man
Caption Game
Click The Button
Zoo Keeper
Add to Google delicious Add to del.icio.us

More ways to get Braingle...
rss


20 Questions: Technical Description

A description of the technology behind the game.

Back to the Game


The 20 Questions Guessing game uses a sophisticated artificial intelligence algorithm to guess the animal that you are thinking about. It uses a neural network to store the answers to all the questions for each animal. Each game that is played modifies the neural network, so the game is constantly getting smarter.

Traditionally, the technology behind this game would be a binary tree where each node of the tree is a question and the leaves of the tree are the animals. You take a different path down the tree based on the answer to each questions. This method has some serious disadvantages. First, this method is intollerant of mistakes by the player. Any questions answered wrong will move the player down the wrong branch of the tree and the correct animal will never be guessed. Second, using the tree method, all questions require a yes or no answer. Many questions about animals are sometimes ambiguous, or can be correctly answers yes and no depending on the interpretation. The method I chose overcomes these two problems.

The 20 Questions game on Braingle.com uses a neural network. Think of this as a matrix where all the rows are animals and all the columns are questions. Each cell represents the answer to one question for one animal. Each answer is represented as a whole number. The more positive the number, the more likely the answer to this questions for this animal is 'yes'. The more negative the number, the more likely the answer is 'no'. Each time an animal is guessed correctly, the neural network is updated to reinforce those questions that were asked.

When you are playing the game, it keeps track of a score for each animal in the database based on the answers that you have given. The higher the score, the more likely the animal is the one you have in your mind. After each question, the game calculates the best possible question to ask you.

If somehow the game quesses incorrectly, it will ask the player to enter the animal's name and a question to differentiate it from the animal it guessed. This information is added to the database so the game is constantly learning new animals and questions.




Online Now: 16 users and 566 guests

Copyright © 1999-2014 | Updates | FAQ | RSS | Widgets | Links | Green | Subscribe | Contact | Privacy | Conditions | Advertise

Custom Search





Sign In A Create a free account