Automatic analysis of socail networks

The threat of rude and abuse on the Internet means that many people avoid expressing themselves and deny seeking different opinions. Internet platforms and services are trying to struggle to effectively contribute conversations, permitting user accounts and communities to limit or completely disable comments. The paper presents a study of the existing approaches for text classification task and advantages of the Deep Learning methods over other baseline machine learning methods. The research focuses on investigating questions concerning the toxic classification of comments from social networks by machine learning techniques. All the techniques are evaluated on the mean column-wise ROC AUC score using a dataset of comments from Wikipedia’s talk page edits provided by Toxic Comment Classification Challenge organizers. The competition is presented on the Kaggle platform. The main feature of the present work is building a model that is capable of effectively detecting different types of toxicity i.e. harassments, threats, abuses, rude statements, and identity-based hate which will help online discussion become more auspicious and respectful.