Insult-detection

Various classifiers are trained with forum posts labelled either neutral or insulting.

Preprocessing methods used

• unicode cleanup
• lemmatization
• stopwords
• TFIDF
• POS tagging

Trained models

• Naive bayes
• SVM
• Random forests

Metrics

• F1
• CLF

Libraries used

• nltk
• sklearn
• panda
• numpy

The notebook is fully documented , conclusion and extra thoughts are included there