TORGO

Overview

TORGO is a specialized database of dysarthric speech designed for evaluating ASR systems on speakers with motor speech disorders. It contains aligned acoustic and articulatory data from speakers with cerebral palsy (CP) or amyotrophic lateral sclerosis (ALS).

Task Description

Task Type: Dysarthric Speech Recognition
Input: Audio recordings from speakers with speech disorders
Output: Transcribed text
Focus: Accessibility and inclusive ASR evaluation

Key Features

Specialized dataset for dysarthric speech
Speakers with cerebral palsy (CP) or ALS
Intelligibility-based subsets (mild, moderate, severe)
3D articulatory feature alignment
Important for accessibility research

Evaluation Notes

Default configuration uses test split
Subsets by intelligibility: mild, moderate, severe
Metrics: CER (Character Error Rate), WER (Word Error Rate), SemScore
Requires jiwer package for CER/WER metrics
Requires jellyfish package for SemScore metric
Supports batch scoring for efficiency

Properties

Property	Value
Benchmark Name	`torgo`
Dataset ID	extraordinarylab/torgo
Paper	N/A
Tags	`Audio`, `SpeechRecognition`
Metrics	`cer`, `wer`, `sem_score`
Default Shots	0-shot
Evaluation Split	`test`

Data Statistics

Metric	Value
Total Samples	5,553
Prompt Length (Mean)	67 chars
Prompt Length (Min/Max)	67 / 67 chars

Per-Subset Statistics:

Subset	Samples	Prompt Mean	Prompt Min	Prompt Max
`mild`	1,479	67	67	67
`moderate`	1,666	67	67	67
`severe`	2,408	67	67	67

Audio Statistics:

Metric	Value
Total Audio Files	5,553
Audio per Sample	min: 1, max: 1, mean: 1
Formats	wav

Sample Example

Subset: mild

{
  "input": [
    {
      "id": "1220f252",
      "content": [
        {
          "text": "Please recognize the speech and only output the recognized content:"
        },
        {
          "audio": "[BASE64_AUDIO: wav, ~89.1KB]",
          "format": "wav"
        }
      ]
    }
  ],
  "target": "FEE",
  "id": 0,
  "group_id": 0,
  "subset_key": "mild",
  "metadata": {
    "transcript": "FEE",
    "intelligibility": "mild",
    "duration": 2.8499999046325684
  }
}

Prompt Template

Prompt Template:

Please recognize the speech and only output the recognized content:

Usage

Using CLI

evalscope eval \
    --model YOUR_MODEL \
    --api-url OPENAI_API_COMPAT_URL \
    --api-key EMPTY_TOKEN \
    --datasets torgo \
    --limit 10  # Remove this line for formal evaluation

Using Python

from evalscope import run_task
from evalscope.config import TaskConfig

task_cfg = TaskConfig(
    model='YOUR_MODEL',
    api_url='OPENAI_API_COMPAT_URL',
    api_key='EMPTY_TOKEN',
    datasets=['torgo'],
    dataset_args={
        'torgo': {
            # subset_list: ['mild', 'moderate', 'severe']  # optional, evaluate specific subsets
        }
    },
    limit=10,  # Remove this line for formal evaluation
)

run_task(task_cfg=task_cfg)

3.3 KiB Raw Blame History