Detecting hate speech is challenging because language constantly evolves, often carrying subtle emotional undertones that can be difficult to interpret. While many existing approaches rely on semantic analysis, they frequently miss these emotional cues and struggle to adapt to different contexts. This paper introduces an enhanced Dual Contrastive Learning (DCL) framework, integrating emotion profiling with fine-tuned language models (BERTweet, RoBERTa, TimeLMs) to address these gaps. By using SHAP (SHapley Additive exPlanations) for interpretability, our approach enhances both detection accuracy and explainability. Experiments on SemEval-2019, Davidson, and a Unified Twitter corpus show that our EmotionDCL models outperform baseline methods, with TimeLMs achieving 81.20% accuracy on SemEval and 94.04% on Davidson. Ablation studies confirm the synergy between contrastive learning and emotion integration, highlighting anger, fear, and anticipation as dominant emotional drivers of hate speech. These findings show the importance of emotion-aware, context-adaptive detection systems and the need for strategies to address class imbalance and linguistic evolution.

Objective: To develop and evaluate EDBERT (Emergency Department Bidirectional Encoder Representations from Transformers), a BERT-based architecture built around a multi-input fusion network that enriches free-text triage notes with the additional context of presenting complaints and patient age to predict Emergency Department (ED) disposition, supporting early decision-making and reduced ED length of stay (LOS). Methods: A retrospective cohort of 570,143 ED presentations to the Royal Melbourne Hospital, Australia, was used. At the core of EDBERT, a fusion network integrates three complementary input sources—triage notes, presenting complaints, and patient age—through learnable weights, with the two free-text inputs encoded by customised 8-layer BERT encoder stacks pre-trained on triage text with Masked Language Modelling (MLM). The resulting architecture is also lightweight, containing 66 million parameters, substantially fewer than the 110 million of BERT-BASE. Eighty percent of the dataset was used for training and 20% for testing. Performance was benchmarked against BERT-Base configurations of varying depth. Results: EDBERT achieved an accuracy of 83.26%, a macro-averaged F1-score of 81.99%, and an area under the receiver operating characteristic curve of 0.91, outperforming all single-input BERT-Base variants, including a depth-matched 8-layer model (82.08%), indicating that the gain derives from the fusion of complementary inputs combined with domain-adaptive pre-training rather than from model capacity. Conclusion: A fusion-based BERT architecture that supplements the triage narrative with presenting complaint and age context predicts ED disposition more accurately than larger single-input BERT models, while requiring substantially less memory and computation, making it practical for deployment in hospitals with limited computing resources.

Objective: To develop and evaluate EDBERT (Emergency Department Bidirectional Encoder Representations from Transformers), a BERT-based architecture built around a multi-input fusion network that enriches free-text triage notes with the additional context of presenting complaints and patient age to predict Emergency Department (ED) disposition, supporting early decision-making and reduced ED length of stay (LOS). Methods: A retrospective cohort of 570,143 ED presentations to the Royal Melbourne Hospital, Australia, was used. At the core of EDBERT, a fusion network integrates three complementary input sources—triage notes, presenting complaints, and patient age—through learnable weights, with the two free-text inputs encoded by customised eight-layer BERT encoder stacks pre-trained on triage text with Masked Language Modelling (MLM). The resulting architecture is also lightweight, containing 66 million parameters, substantially fewer than the 110 million of BERT-BASE. Eighty percent of the dataset was used for training, and twenty percent for testing. Performance was benchmarked against BERT-Base configurations of varying depth. Results: EDBERT achieved an accuracy of 83.26%, a macro-averaged F1-score of 81.99%, and an area under the receiver operating characteristic curve of 0.91, outperforming all single-input BERT-Base variants, including a depth-matched eight-layer model (82.08%), indicating that the gain derives from the fusion of complementary inputs combined with domain-adaptive pre-training rather than from model capacity. Conclusions: A fusion-based BERT architecture that supplements the triage narrative with presenting complaint and age context predicts ED disposition more accurately than larger single-input BERT models, while requiring substantially less memory and computation, making it practical for deployment in hospitals with limited computing resources.