Probability Project

This project implements a Bayesian text classifier that estimates whether a given text is more likely written in English or Spanish, based on character frequency distributions and multinomial probability models.

Overview

The classifier uses:

• Case-folding (ignoring letter case)
• Bag-of-characters counts (A–Z)
• Log-probability calculations to avoid numerical underflow
• Bayesian inference to compute the posterior probability

Key Steps

1. Load probability vectors for English and Spanish
2. Count characters in the input file
3. Compute log-likelihoods for both languages
4. Apply Bayes’ rule using a logistic function
5. Output the predicted probability

How to Run

Run with default priors (0.6 English, 0.4 Spanish):

python3 probability.py letter0.txt

Run with custom priors:

python3 probability.py letter0.txt 0.5 0.5

Files

• probability.py — main script
• e.txt — English character probability vector
• s.txt — Spanish character probability vector
• letter*.txt — sample input files
• letter*_out.txt — generated output files

Techniques Used

• Multinomial probability modeling
• Log-sum calculations
• Bag-of-characters feature extraction
• Bayesian classification logic
• Python-based text processing

Author

Macy Xiang
GitHub: https://github.com/macyxiangA