Ανίχνευση κυβερνοεκφοβισμού στις αναρτήσεις χρηστών με την χρήση του μοντέλου BERT
Cyberbullying detection on social networks using bert model
View/ Open
Keywords
Βαθιά μάθηση ; Ταξινόμηση κειμένου ; Μετασχηματιστές ; Μοντέλο BERT ; Fine-tuning ; Επεξεργασία φυσικής γλώσσας ; Κοινωνικά δίκτυαAbstract
Nowadays, social networks have a vital influence on our lives as they are a daily and necessary means of communication and means of expression for the majority of people. However, there are some users who use them with the intent to insult, harass, deceive and threaten other users. Cyberbullying is a phenomenon that includes incidents like the ones we mentioned earlier. In order to protect users of social networking platforms from this problem, it is considered necessary to find ways to prevent such attacks. This is achieved if cyberbullying can be detected in users messages. Text classification is an important and useful task in the natural language processing field, with a wide range of applications. So, the objective of this thesis is the creation of a classifier that recognizes instances of text, written in the English language, in which cyberbullying occurs. In order to achieve this, we used Google's BERT model.