BERT-based ensemble learning for multi-aspect hate speech detection

Mazari, Ahmed Cherif; Boudoukhani, Nesrine; Djeffal, Abdelhamid

doi:10.1007/s10586-022-03956-x

Cited by 24 publications

(5 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This multifaceted landscape of research showcases the dynamic and evolving nature of hate speech detection methodologies, each contributing valuable insights to the overarching pursuit of fostering safer online environments. [8] The study addresses the pervasive issue of unfiltered content on social media by proposing a multi-aspect hate speech detection approach. Leveraging pre-trained BERT models and combining them with Deep Learning models, including Bidirectional LSTM and Bidirectional GRU on GloVe and FastText word embeddings, the researchers achieve a notable 98.63% ROC-AUC score in enhancing hate speech detection on social media, demonstrating the efficacy of their comprehensive ensemble learning strategy.…”

Section: Related Workmentioning

confidence: 99%

Innovative Fusion of LSTM and Bi-GRU Networks for Enhanced Hate Speech Detection in Social Media

2024

IRJMETS

View full text Add to dashboard Cite

In recent times, the escalating prevalence of hate speech on various social media platforms has emerged as a pressing concern, underscoring the critical need for the development of resilient and effective mechanisms aimed at automated detection. This imperative is not only rooted in the imperative to combat the proliferation of hate speech but also in the broader goal of enhancing the overall user experience within online communities. Against this backdrop, this research endeavors to present an innovative and sophisticated approach to hate speech detection, employing the capabilities of combination of Long Short-Term Memory (LSTM) networks and Bidirectional Gated Recurrent Unit (Bi-GRU)The assessment of the proposed hate speech detection model is meticulously conducted through the utilization of standard performance metrics, including precision, recall, and F1 score. The outcomes of these analyses reveal a noteworthy enhancement in the model's accuracy in discerning instances of hate speech, thereby attesting to the substantial potential of LSTM-BiGRU networks in effectively addressing the multifaceted challenges posed by the ever-evolving and context-dependent nature of hate speech. This research not only contributes to the technological landscape of automated hate speech detection but also underscores the broader societal significance of fostering online environments that prioritize safety, inclusivity, and a positive user experience.

show abstract

Section: Related Workmentioning

confidence: 99%

Innovative Fusion of LSTM and Bi-GRU Networks for Enhanced Hate Speech Detection in Social Media

2024

IRJMETS

View full text Add to dashboard Cite

show abstract

“…One notable work (Mazari et al, 2023 ) presented a novel approach to multi-aspect hate speech detection by leveraging ensemble learning techniques that combined BERT with Deep Learning models like Bi-LSTMs and Bi-GRUs. These models were trained individually and then their outputs were combined to improve the precision of hate speech detection, particularly in identifying various forms of toxic language such as “identity hate”, “threat”, “insult”, “obscene”, “toxic”, and “severe toxic”.…”

Section: Related Workmentioning

confidence: 99%

Hate speech detection with ADHAR: a multi-dialectal hate speech corpus in Arabic

Charfi,

Besghaier,

Akasheh

et al. 2024

Front. Artif. Intell.

View full text Add to dashboard Cite

Hate speech detection in Arabic poses a complex challenge due to the dialectal diversity across the Arab world. Most existing hate speech datasets for Arabic cover only one dialect or one hate speech category. They also lack balance across dialects, topics, and hate/non-hate classes. In this paper, we address this gap by presenting ADHAR—a comprehensive multi-dialect, multi-category hate speech corpus for Arabic. ADHAR contains 70,369 words and spans four language variants: Modern Standard Arabic (MSA), Egyptian, Levantine, Gulf and Maghrebi. It covers four key hate speech categories: nationality, religion, ethnicity, and race. A major contribution is that ADHAR is carefully curated to maintain balance across dialects, categories, and hate/non-hate classes to enable unbiased dataset evaluation. We describe the systematic data collection methodology, followed by a rigorous annotation process involving multiple annotators per dialect. Extensive qualitative and quantitative analyses demonstrate the quality and usefulness of ADHAR. Our experiments with various classical and deep learning models demonstrate that our dataset enables the development of robust hate speech classifiers for Arabic, achieving accuracy and F1-scores of up to 90% for hate speech detection and up to 92% for category detection. When trained with Arabert, we achieved an accuracy and F1-score of 94% for hate speech detection, as well as 95% for the category detection.

show abstract

“…In numerous studies, BERT outperformed conventional machine learning algorithms, achieved cutting-edge performance, and demonstrated promising results in the detection of cyberbullying. Studies [22,[58][59][60] show that BERT achieved high accuracy and F1 scores to classify cyberbullying in various types of online content, such as tweets and comments. The performance of BERT may vary depending on the dataset and task.…”

Section: Fasttextmentioning

confidence: 99%

“…Using a pre-trained BERT model along with deep learning (DL) models, Mazari et al [58] proposed a multi-aspect hate speech detection approach based on text classification in multiple labels. Bidirectional Long-Short Term Memory (Bi-LSTM) and/or Bidirectional Gated Recurrent Unit (Bi-GRU) are stacked on GloVe and FastText word embeddings to create the DL models that are used.…”

Section: Fasttextmentioning

confidence: 99%

A Review on Deep-Learning-Based Cyberbullying Detection

et al. 2023

View full text Add to dashboard Cite

Bullying is described as an undesirable behavior by others that harms an individual physically, mentally, or socially. Cyberbullying is a virtual form (e.g., textual or image) of bullying or harassment, also known as online bullying. Cyberbullying detection is a pressing need in today’s world, as the prevalence of cyberbullying is continually growing, resulting in mental health issues. Conventional machine learning models were previously used to identify cyberbullying. However, current research demonstrates that deep learning surpasses traditional machine learning algorithms in identifying cyberbullying for several reasons, including handling extensive data, efficiently classifying text and images, extracting features automatically through hidden layers, and many others. This paper reviews the existing surveys and identifies the gaps in those studies. We also present a deep-learning-based defense ecosystem for cyberbullying detection, including data representation techniques and different deep-learning-based models and frameworks. We have critically analyzed the existing DL-based cyberbullying detection techniques and identified their significant contributions and the future research directions they have presented. We have also summarized the datasets being used, including the DL architecture being used and the tasks that are accomplished for each dataset. Finally, several challenges faced by the existing researchers and the open issues to be addressed in the future have been presented.

show abstract

BERT-based ensemble learning for multi-aspect hate speech detection

Cited by 24 publications

References 40 publications

Innovative Fusion of LSTM and Bi-GRU Networks for Enhanced Hate Speech Detection in Social Media

Innovative Fusion of LSTM and Bi-GRU Networks for Enhanced Hate Speech Detection in Social Media

Hate speech detection with ADHAR: a multi-dialectal hate speech corpus in Arabic

A Review on Deep-Learning-Based Cyberbullying Detection

Contact Info

Product

Resources

About