Datasets:

Kppwdfgu1
/

4d69448f855bde30

Tasks:

Automatic Speech Recognition

Modalities:

Formats:

Size:

Libraries:

Dataset card Data Studio Files Files and versions

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

4d69448f855bde30

Before/after timestamp-correction comparison

Each row contains one unchanged SBPN transcript and two lossless audio cuts:

audio_before: original transcript_sbpn timestamps
audio_after: corrected transcript_sbpn_diarized timestamps
text: the unchanged transcript, shown once

Only rows whose timestamps changed are included.

Examples: 741
Recordings represented: 26
Start-only corrections: 346
End-only corrections: 344
Both boundaries corrected: 51
Before audio: 0.541 hours
After audio: 0.577 hours
Boundary audio added: 130.727 seconds

Speaker labels are anonymous and recording-local. This dataset is intended for manual comparison and alignment QA. Confirm rights, consent, and permitted use before redistribution or training.

Downloads last month: 28

Total file size: