In our contribution, we have applied stance analysis in order to identify offensive discourse. This gives us access to the pros and cons of the writer of some tweets and reveals his/her role framing of the discourse referents. We also semi-automatically augmented our polarity lexicon with a new type of polarity labels, namely P for profanity. Starting from seed words, we derived new entries on the basis of word embeddings. Our approach also focuses on offensive language without offensive words (OLWOW) and discusses strategies to cope with it.