Welcome to the 19 Questions game!

19 Questions is a machine learning game which asks you questions and guesses an object you are thinking about. It uses your answers to improve its interrogation technique.

Access the demo site at: https://phor.net/apps/19q/

Although inspired by neural networking, 19 Questions is based entirely on the Bayesian inference technique.

Run php -S localhost:8888 from inside this folder and point your web browser to http://localhost:8888.

It uses SQLite so the database is already set up and checked into this repository.

For production use, copy all files to your web server. Also consider using a different database (see SCHEMA.sql) for installations with more than 10,000 visitors per hour.

This program remembers past games and uses this as evidence to assume connections between objects (e.g. “microwave") and predicates (e.g. “is larger than a loaf of bread”).

The program uses statistical inference (unbiased maximum likelihood estimator) to estimate which object you are thinking of based on answers you provide. Then it asks you a question which hopefully will bifurcate the remaining universe of likely objects.

The computer wins if it correctly guesses the player’s object within 19 questions.

For more discussion of the algorithm, see TECHNICAL.md.

William Entriken is interested to learn more about machine learning and helps SEPTA, Philadelphia's local municipal transport system, improve their rail schedules. Read more about the author at: https://phor.net/.

Contributions to 19 Questions are welcome and encouraged!

If you can improve the statistical calculations, presentation of information or technical documentation, this would be most appreciated.