Datasets:

TalhaAhmed
/

gal-fe01-clean

Tasks:

Automatic Speech Recognition

Modalities:

Formats:

Languages:

Size:

Tags:

Libraries:

License:

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.

TalhaAhmed/gal-fe01-clean

Cleaned version of TalhaAhmed/gal-fe01 Generated on 2026-03-24 11:14 UTC by clean_urdu_tts_dataset.py.

Cleaning summary

Issue type	Rows affected	Action
Word / phrase repetition	1,868 (+ 1,547 phrase)	First occurrence kept
Character-level stutter	9	Collapsed to 2 chars
Noise / filler tokens	194	Removed
Stray ASCII words	0	Stripped
Garbage / non-Urdu rows	0	Dropped
Too short after cleaning	717	Dropped

Row counts

	Count
Original rows	28,846
Rows kept	28,099
Rows dropped	747 (2.6%)
Rows with text fixed (kept)	1,397

Schema

Identical to the source dataset. Only the text (transcript) column is modified. The audio column and all other columns are untouched.

A full per-row audit trail is available in cleaning_report.csv in this repo.

Source

<TalhaAhmed/gal-fe01> — TalhaAhmed/gal-fe01

Downloads last month: 20