Ablation Experiment Results for Emotion Recognition Model Performance

Weiguang Dong's ablation experiment results, uploaded to figshare on 2026-04-27. The 5.5 KB XLS file contains performance metrics for a hybrid emotion recognition model evaluated on three benchmark datasets: SemEval-2018, RAVDESS, and CMU-MOSEI. The model achieved state-of-the-art results, including 58.5% accuracy on SemEval-2018 and 82.3% accuracy on CMU-MOSEI.

Use Cases

• Benchmarking new emotion recognition models against reported accuracy and F1-scores.
• Analyzing the impact of model components (E-GAT and Bi-LSTM) based on ablation study results.
• Comparing model performance across text-only and multimodal datasets (SemEval-2018, RAVDESS, CMU-MOSEI).

Strengths

• Performance metrics are provided for three established benchmark datasets.
• The dataset is small (5.5 KB) and in a widely accessible XLS format.
• Results include specific accuracy scores (e.g., 58.5%, 78.9%, 82.3%) and an F1-score (68.7%).

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• The dataset's tiny scale (5.5 KB) indicates it contains only summary results, not the underlying training or evaluation data.

Provenance

Source

Weiguang Dong

Collection Method

Results from ablation experiments for a proposed hybrid emotion recognition model.

Time Range

The data reflects experiments conducted prior to the 2026-04-27 upload date.

Freshness

Last updated 2026-04-27 17:22:32; freshness should be verified.

Geography

null