University of Twente Student Theses
A Comparative study of BERT-CNN and GCN for Hate Speech Detection
Pandey, S.S. (2023) A Comparative study of BERT-CNN and GCN for Hate Speech Detection.
|
PDF
776kB |
| Abstract: | Social media has become not only a medium for like-minded people to connect but also a platform where anyone can freely express their thoughts and opinions. However, its widespread nature has not only led to an immeasurable impact on society but has also presented some important challenges. Online hate speech is one such challenge. Consequently, the identification of hate speech on online platforms has gained much traction recently. Different methods ranging from reactive approaches like using Natural Language Processing (NLP) for classifying individual posts to proactive approaches like using contextual information and predicting when a discussion advances towards hatefulness have been tried in the domain of Hate Speech Detection. In this paper, we perform an in-depth comparison of two such techniques of Hate Speech Classification, namely BERT-CNN and Graph-based Graph Convolutional Network (GCN). Our findings show that when developed on the same dataset from Twitter, the BERT-CNN model requires fewer computational resources compared to the GCN model. Moreover, the BERT-CNN model achieves a macro F1 score of 0.81 outperforming the GCN model with a macro F1 score of 0.48. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/95839 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page