This project implements the multinomial Naive Bayes with Nigram's many-to-one assumption EM algorithm (semi-supervised learning).
The implement of EM algorithm derives from the basic multinomial Naive Bayes
of scikit-learn's MultinomialNB
.
- Install requirements.txt
- All test cases are implemented in
source.py
as individual functions.
data/...
all data processing scriptsmultinomial_model
implements model,utilities
supports some simple graph construction methods