Common Voice Scripted Speech 24.0 - Pashto Integration Guide

Common Voice Scripted Speech 24.0 - Pashto Integration Guide

This project recognizes Mozilla Common Voice as a major source for Pashto ASR progress and community participation.

Dataset

• Name: Common Voice Scripted Speech 24.0 - Pashto
• Dataset page: Mozilla Data Collective - Common Voice Pashto 24.0
• Release date: 2025-12-05
• Format: MP3 with TSV metadata
• Approximate size: 49.98 GB
• License: CC0-1.0

Important Usage Rules

• Do not attempt to identify speakers.
• Do not re-host or re-share the raw dataset files.
• Keep provenance and version information when reporting experiments.

How To Use In This Repository

1. Download from the official Mozilla Data Collective page.
2. Extract locally under: data/raw/common_voice_scripted_ps_v24/
3. Keep raw audio out of git.
4. Use project scripts/docs for normalization, splits, and benchmarking.

Recommended local structure:

data/raw/common_voice_scripted_ps_v24/
  clips/
  train.tsv
  dev.tsv
  test.tsv

How To Contribute Through Mozilla Common Voice

Contributors can directly improve Pashto resources on Common Voice:

• Speak: commonvoice.mozilla.org/ps/speak
• Write: commonvoice.mozilla.org/ps/write
• Listen: commonvoice.mozilla.org/ps/listen
• Review: commonvoice.mozilla.org/ps/review

Contribution Loop Back To This Project

After contributing on Common Voice, open an issue/PR here and share:

• what task you worked on (speak/write/listen/review),
• what quality gaps you observed,
• what dataset or modeling step should be improved next.

Use issue labels:

• data
• good first issue
• help wanted