Confidential — Stefan Michaelcheck Only

BERT-OTA: Enhancing Hate Speech Detection With Ontology-Guided Transformer Attention

2026model innovationincrementalmethod

Mahmoud Abusaqer, Mukulika Ghosh, Jamil Saquer

IEEE Access

https://doi.org/10.1109/access.2026.3650874OpenAlex: W7118162908
1
URLs Found
0
Internal Citations
3
Authors
usable
Abstract Quality
GPT-5.5 Abstract Analysis

Problems Identified (4)

Hate speech detection challenge: Social media hate speech creates a significant content moderation challenge requiring methods that understand explicit and implicit harmful content.

Contextual nuance limitations: Traditional hate speech detection approaches struggle with contextual nuances and evolving online hate speech patterns.

Hate speech detection challenge: Social media hate speech creates a significant content moderation challenge requiring methods that understand explicit and implicit harmful content.

Contextual nuance limitations: Traditional hate speech detection approaches struggle with contextual nuances and evolving online hate speech patterns.

Proposed Solutions (4)

BERT-OTA: The paper proposes BERT-OTA, an ontology-guided transformer attention architecture for hate speech detection.

Dual-stream BERT-GCN architecture: The approach processes text with BERT and scaled dot-product attention while learning ontological features with a two-layer Graph Convolutional Network.

BERT-OTA: The paper proposes BERT-OTA, an ontology-guided transformer attention architecture for hate speech detection.

Dual-stream BERT-GCN architecture: The approach processes text with BERT and scaled dot-product attention while learning ontological features with a two-layer Graph Convolutional Network.

Results (3)

State-of-the-art hate speech detection performance:

Structured knowledge improves detection:

State-of-the-art hate speech detection performance:

Research Domain

Hate speech detection / online content moderation

← Back to all papers