Toxic online content has end up a critical difficulty in nowadays’s world due to partner in Nursing exponential boom in the use of internet by means of parents of various cultures and academic history. Differentiating hate speech and offensive language can be a key undertaking in automatic detection of virulent text content. throughout this paper, we have a tendency to propose partner in Nursing method to mechanically classify tweets on Twitter into 3 instructions: hateful, offensive and easy. Victimization Twitter dataset, we have a tendency to carry out experiments thinking about n-grams as alternatives and spending their term frequency-inverse document frequency (TFIDF) values to a couple of system getting to know models. We tend to perform comparative evaluation of the models considering many values of n in n-grams and TFIDF normalization techniques. when standardization the version giving the most effective effects, we have a tendency to accomplish ninety five.6% accuracy upon evaluating it on take a look at expertise. we tend to conjointly produce a module that is partner in Nursing intermediate between user and Twitter.