Hate Speech Detection Leveraging Emoji Descriptions and BERT-Based Contextual Data Augmentation
ComTech: Computer, Mathematics and Engineering Applications · 28.09.2026 · 17 (2) · 117-125
DOI: 10.21512/comtech.v17i2.14618 ↗Kurz gefasst
Fachartikel in der Zeitschrift ComTech: Computer, Mathematics and Engineering Applications, online erschienen am 28.09.2026. Verfasst von Junita Amalia und 3 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
This study evaluates the efficacy of emoji-to-text transformation and contextual data augmentation in improving hate speech detection on social media. Using a unified dataset from OLID, TweetEval, and the Sushil benchmark, we compared Multinomial Naive Bayes (MNB), Bidirectional LSTM (Bi-LSTM), and Multilingual BERT (mBERT). To address class imbalance and semantic loss, we implemented a preprocessing pipeline that converts emojis into textual descriptions and utilized a BERT-based contextual augmenter. Results indicate that textualizing emojis consistently improves performance by providing critical semantic anchors. The mBERT model achieved the highest performance with a Macro F1-score of 0.87 and a hate speech recall of 0.94, outperforming the Bi-LSTM (0.835) and MNB (0.83) baselines. Notably, the Bi-LSTM showed the highest sensitivity to the "Emoji Description" method, with a 5% performance increase compared to the "No Emoji" condition. These findings demonstrate that bridging the gap between visual symbols and textual intent significantly reduces model bias and improves the detection of nuanced toxicity. This research provides an optimized framework for context-aware moderation in imbalanced digital environments.
Zitieren
APA
Amalia, J., Simanjuntak, W. V., Hutagalung, R. M., & Tampubolon, L. M. (2026). Hate Speech Detection Leveraging Emoji Descriptions and BERT-Based Contextual Data Augmentation. ComTech: Computer, Mathematics and Engineering Applications, 17(2), 117-125. https://doi.org/10.21512/comtech.v17i2.14618
Vancouver
Amalia J, Simanjuntak WV, Hutagalung RM, Tampubolon LM. Hate Speech Detection Leveraging Emoji Descriptions and BERT-Based Contextual Data Augmentation. ComTech: Computer, Mathematics and Engineering Applications. 2026;17(2):117-125. doi:10.21512/comtech.v17i2.14618
BibTeX
@article{Amalia2026,
author = {Amalia, Junita and Simanjuntak, Walker Valentinus and Hutagalung, Ruth Marelisa and Tampubolon, Lamria Magdalena},
title = {Hate Speech Detection Leveraging Emoji Descriptions and BERT-Based Contextual Data Augmentation},
journal = {ComTech: Computer, Mathematics and Engineering Applications},
year = {2026},
volume = {17},
number = {2},
pages = {117--125},
doi = {10.21512/comtech.v17i2.14618}
}
EUVIMED