Νευρωνικά δίκτυα Transformer για την ταξινόμηση συναισθήματος
Transformer neural networks for sentiment classification
View/
Keywords
Νευρωνικά δίκτυα ; Transformer ; Ταξινόμηση συναισθήματοςAbstract
This thesis shows how Transformer neural networks can be used for sentiment classification. Specifically, the pre-trained BERT model which is based on the Transformer architecture for linguistic modelling is used to perform sentiment analysis on the GoEmotions dataset. Firstly, the reasons why sentiment classification is important and the research that has been done on it are discussed. Then, the architectures of basic types of neural networks, the transformer model and the BERT model are shown and a description of the GoEmotions dataset is given. Afterwards, the results of the classification are presented and its evaluation through the appropriate metrics is done. Finally, the conclusions drawn from the classification are listed and suggestions for future research are given.
Postgraduate Studies Programme
Ψηφιακός Πολιτισμός, Έξυπνες Πόλεις, IoT και Προηγμένες Ψηφιακές ΤεχνολογίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
62Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα