Datasets:

MergenAI
/

YouTube-Labeled-full

Parquet error: Scan size limit exceeded: attempted to read 906825226 bytes, limit is 300000000 bytes Make sure that 1. the Parquet files contain a page index to enable random access without loading entire row groups2. otherwise use smaller row-group sizes when serializing the Parquet files

Error code:   TooBigContentError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YouTube Azerbaijani Segment-Level ASR Dataset (v0.8) — Parquet

Overview

This dataset contains segment-level Azerbaijani audio/transcript pairs stored as Parquet shards. Each row includes the audio bytes (16 kHz, mono, 16-bit PCM WAV) and full metadata.

Source audio: YouTube Azerbaijani audio (checkpointed feed)
Labels: vox-labeler labels.jsonl segments
Audio format: 16 kHz, mono, 16-bit PCM WAV
Repo: MergenAI/YouTube-Labeled-full
Generated at: 2026-01-27 13:41 UTC

Dataset Structure

README.md
data/
  train/
    data-00000-of-XXXXX.parquet
    data-00001-of-XXXXX.parquet
    ...

Parquet Schema (high-level)

audio (struct): bytes, path, sampling_rate
text (string)
file_name (string)
audio_id (string)
segment_index (int)
segment_start (float)
segment_end (float)
segment_duration (float)
audio_duration_s (float)
source_duration_s (float|null)
source_sample_rate (int|null)
segment_avg_logprob (float|null)
segments_total (int)
label_text (string)
label_avg_logprob (float|null)
label_no_speech_prob (float|null)
language (string|null)
model_version (string)
run_id (string)
decode_params_json (string|null)
filters_json (string|null)
source_part (string)
source_audio_relpath (string)

Sharding

Rows per shard (target): 1000
Max shard bytes (target): 1.00 GB
Compression: snappy

Stats

Total rows: 752540
Total audio duration (segments): 5893.81 hours (353628.87 minutes)
Total shards: 753

Usage (datasets)

from datasets import load_dataset, Audio

ds = load_dataset("MergenAI/YouTube-Labeled-v0.8", split="train")
ds = ds.cast_column("audio", Audio())
print(ds[0])

Downloads last month: 33

Size of downloaded dataset files:

670 GB

Size of the auto-converted Parquet files:

670 GB

Number of rows:

752,540