The purpose of this to develop models to identify and categorize offensive language. It consists in three subtasks:
- classify the tweets as being offensive or not
- focus only on the offensive tweets and distinguish tweets which target somebody or something, and tweet which are not targeted
- determine whether the offensive tweets are targeting individuals, groups or ”other” categories which could be events, organisations...
In this notebook, I described the different steps to tackle this NLP challenge: Data augmentation, Data preprocessing, Data modeling (via Recurrent Neural Networks), ...