Εφαρμογές Γλωσσικής Τεχνολογίας στα Ελληνικά για την Αναγνώριση Προτύπων Μάθησης και Συμπεριφοράς στις Δυνητικές Κοινότητες Μάθησης: Σχολικός Εκφοβισμός και Αυτοκτονικός Ιδεασμός
Date Issued
June 29, 2020
Type
Διδακτορική Διατριβή
Abstract
Aggressive behavior has become a major social problem concerning sensitive groups, such as children and adolescents. It occurs mainly in two different types: aggression "towards the other" and aggression "towards the self". The first type refers mainly to bullying, while the second one refers to suicidal ideation. Early detection of both types of aggressive behavior is crucial for timely intervention.
Nowadays, various activities are transferred from real world to the digital one. Consequently, bullying acquires its digital form, cyberbullying. Accordingly, suicidal behavior is transformed into cyberbullicide. At the same time, online collaborative learning environments have been developed, transferring learning processes to the digital world. As a result, transition of Physical Learning Communities (PLCs) to the digital world, as Virtual Learning Communities (VLCs), includes the risk of aggressive behavior.
The development of Computational Linguistics' applications has led to the creation of models, aiming at the automatic detection of aggressive behavior, especially in the digital world, where physical supervision is particularly difficult. Research so far has provided promising results, but it is limited to the social media context and it has not been applied to the Greek language.
In the present dissertation, the aggressive behavior in the field of VLCs has been studied and analyzed. For this purpose, five (5) case studies (CS) were created, which were designed methodologically as a collective CS. An analysis, based on the collaborative process, the role of the teacher and the inner speech, was applied on the collected data (dialogues and artifacts). Suicidal ideation was also analyzed using linguistic data drawn from modern Greek poetry.
Models, aiming to identify both types of aggressive behavior using Natural Language Processing (NLP) and machine learning methods, were applied to linguistic data of different types, providing promising results.
Application of these models in VLCs' data and in the Greek language is the main contribution of this dissertation.
Subjects