This project implements a simple Shannon game to estimate the entropy of human language models, for my NLP class (Fall 2022, Oregon State).

It was inspired by the Java Applet from UCSD's Math 187 class (Cryptography, Spring 2014) by Prof. Adriano Garsia. Since Java Applets have become obsolete, I wrote this very simple Python tool as a replacement for my students. It uses ANSI escape sequences for cursor movement.

This package uses the readchar library.

pip3 install -r requirements.txt

On a Linux/Mac terminal (better with wide screen), run

python3 shannon_game.py [<text_file>]

If the argument is missing, test.txt is used by default.

A typical screenshot is:

Welcome to the Shannon Game! Please guess!
118 chars remaining
EVEN THOUGH YOU _
2           2
01121413411161117
[-BCD-FGHIJKL-NOPQ--TUV-XYZ-]
entropy in range [1.44, 1.80]

This project uses Shannon's original estimates of the lowerbound and the upperbound, but using N=1 (i.e., unigram guess sequence).

Liang Huang ([email protected])