Twenty questions

When I was doing my research I found a couple paragraphs written by the wife of the inventor of the game 20 question. These couple paragraghs that she wrote really summed everything up and got me started on my research, so here they are:
Mr. Burgener, based in Ottawa, Canada, first developed the software in 1988, to satisfy his own curiosity about learning systems. "The programmatic solution seems fairly simple," he starts out to say. "Build a tree. If the program fails, ask for a question to differentiate the last object from the new object. But, if one player answers a question from a different perspective, the game will fail; player input is subject to human nature and cannot be 100 per cent correct every time."
"The solution I have used," he said, "relies on a matrix of every object by every question; each element in the matrix indicates an answer for one object and one question." In very simple terms, the program determines the most probable objects and then looks at the questions to find the question that could eliminate the most objects. "But then it gets fuzzy," says Mr. Burgener. "The answers in the knowledgebase are weighted by certainty, and are, more often than not, unknown rather than known."
"a failure for the game is not really a failure, because the program learns from every game that is played."
Sourcehttp://www.mikiko.net/library/weekly/1998articles/aa090698.htm
I picked to do my project on this topic because I was always so amazed on how this game works and I had a craving to answer my question. How does a battery powered, $20 toy have the mind of a computer? And that is exactly what I'm going to figure out. I first developed a love for this game when I gave 20 questions out for a goody-bag in 4^th grade. Now I play it whenever I can't go to sleep and I never stopped wondering how it works.
Here is a video link that explains how it works in detail. This will be my main source. Please watch it you will be amazed and it is funny ! J

http://www.sittingo.com/talk/713
By,
CK