yassersouri/classify-text

Implements comparative experiments across multiple feature representations (Bag of Words, TF, TF-IDF) and classifiers (Naive Bayes, SVM, k-NN) using scikit-learn, with evaluation via train-test splits and stratified k-fold cross-validation. Handles preprocessing quirks like UTF-8 incompatibility in source documents and supports both binary classification (likes vs. dislikes) and full 20-class multiclass scenarios. Results demonstrate TF-IDF with linear SVM achieving ~97% accuracy on binary tasks and ~89% on full 20-class classification.

147 stars. No commits in the last 6 months.