Trelis/eval-whisper-v3-multimed-hard-20260408-1936
收藏Hugging Face2026-04-08 更新2026-04-12 收录
下载链接:
https://hf-mirror.com/datasets/Trelis/eval-whisper-v3-multimed-hard-20260408-1936
下载链接
链接失效反馈官方服务:
资源简介:
---
tags:
- whisper
- evaluation
- speech
- speech-to-text
---
# Evaluation Results: whisper-v3
Evaluation results from Whisper model evaluation.
## Summary
| Model | WER | CER |
|-------|-----|-----|
| [fireworks/whisper-v3](https://huggingface.co/fireworks/whisper-v3) | 12.95% | 8.99% |
## Source Data
- **Evaluation Dataset:** [Trelis/multimed-hard](https://huggingface.co/datasets/Trelis/multimed-hard)
- **Model Evaluated:** [fireworks/whisper-v3](https://huggingface.co/fireworks/whisper-v3)
## Columns
| Column | Description |
|--------|-------------|
| `audio` | Audio sample (if available from source dataset) |
| `reference` | Ground truth transcription |
| `prediction` | Model prediction |
| `wer` | Word Error Rate for this sample |
| `cer` | Character Error Rate for this sample |
| `entities` | Entity annotations from source dataset |
| `entity_cer` | Per-sample entity CER (-1.0 if no entities) |
## Entity CER
**Overall Entity CER:** 26.07%
| Category | CER |
|----------|-----|
| anatomy | 28.94% |
| biomarker | 19.58% |
| condition | 28.81% |
| drug | 7.14% |
| organisation | 30.69% |
| procedure | 26.04% |
---
*Generated by [Trelis Studio](https://studio.trelis.com)*
标签:
- Whisper(Whisper)
- 模型评测
- 语音
- 语音转文字
# 评测结果:Whisper-v3模型
本数据集收录Whisper模型的评测结果。
## 评测摘要
| 模型 | 词错误率(Word Error Rate,WER) | 字符错误率(Character Error Rate,CER) |
|-------|-----|-----|
| [fireworks/whisper-v3](https://huggingface.co/fireworks/whisper-v3) | 12.95% | 8.99% |
## 源数据集信息
- **评测数据集**:[Trelis/multimed-hard](https://huggingface.co/datasets/Trelis/multimed-hard)
- **待评测模型**:[fireworks/whisper-v3](https://huggingface.co/fireworks/whisper-v3)
## 数据列说明
| 列名 | 说明 |
|--------|-------------|
| `audio` | 音频样本(若源数据集提供该字段) |
| `reference` | 标准转录真值标注 |
| `prediction` | 模型输出的预测转录结果 |
| `wer` | 当前样本的词错误率 |
| `cer` | 当前样本的字符错误率 |
| 列名 | 说明 |
|--------|-------------|
| `entities` | 源数据集自带的实体标注信息 |
| `entity_cer` | 当前样本的实体字符错误率(若无实体则取值为-1.0) |
## 实体字符错误率
**整体实体字符错误率:26.07%**
| 类别 | 字符错误率(CER) |
|----------|-----|
| 解剖学(anatomy) | 28.94% |
| 生物标志物(biomarker) | 19.58% |
| 病症(condition) | 28.81% |
| 药物(drug) | 7.14% |
| 组织机构(organisation) | 30.69% |
| 诊疗操作(procedure) | 26.04% |
*由 [Trelis Studio](https://studio.trelis.com) 生成*
提供机构:
Trelis


