Eval Parakeet Tdt 0.6B V3: Medical Terms Speech Recognition Evaluation

Evaluation results for the NVIDIA Parakeet-tdt-0.6b-v3 speech recognition model on a dataset of medical terms. The dataset contains model predictions, ground truth transcriptions, and error metrics like Word Error Rate (WER) and Character Error Rate (CER). It was created by Trelis and last updated on April 8, 2026.

Use Cases

• Benchmarking speech recognition model accuracy based on Word Error Rate (WER) and Character Error Rate (CER) metrics.
• Analyzing model performance on specialized medical vocabulary based on the reference and prediction columns.
• Comparing model outputs against ground truth transcriptions for error analysis.

Strengths

• Includes specific error metrics: 11.34% Word Error Rate (WER) and 3.63% Character Error Rate (CER) for the evaluated model.
• Provides a direct comparison between model predictions and ground truth transcriptions for medical terms.

Limitations

• Row count, file formats, and data size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Trelis, via Hugging Face.

Collection Method

Likely contains evaluation results from a Whisper model evaluation framework.

Time Range

2025 (from source dataset name 'medical-terms-2025').

Freshness

Last updated 2026-04-08 19:27:10; freshness should be verified.

Geography

null