Toxic comments classification

This dataset comprise of Twitter comments which have been labeled by human raters for toxic behavior. The types of toxicity are:

• toxic
• severe_toxic
• obscene
• threat
• insult
• identity_hate

File descriptions

• train.csv: the training set, contains comments with their binary labels.
• test.csv: the test set, you must predict the toxicity probabilities for these comments. To deter hand labeling, the test set contains some comments which are not included in scoring.
• sample_submission.csv: a sample submission file in the correct format.
• test_labels.csv: labels for the test data; value of -1 indicates it was not used for scoring.