khursanirevo/multiturn_ks

khursanirevo/multiturn_ks

Dataset Description

Multiturn dialogue dataset with speaker-separated stereo audio and multi-language transcripts from 139 YouTube videos.

Features

• —Audio: Stereo audio with speaker separation (speaker 0 = left channel, speaker 1 = right channel)
• —Segments: Speaker turn-level annotations with timestamps for English and Malay
• —Multi-language: Transcripts in 9 languages (en, ms, zh-Hans, zh-Hant, ru, id, ar, ja, ko)
• —Video ID: YouTube video identifier for each chunk
• —Chunking: 30-second chunks with 0.5s overlap

Columns

• —audio: Playable stereo audio (24kHz)
• —video_id: YouTube video identifier
• —sentence: Full transcript for the chunk (English)
• —segments_en: JSON list of English speaker turns [{speaker, start, end, text}]
• —segments_ms: JSON list of Malay speaker turns [{speaker, start, end, text}]
• —total_speakers: Number of speakers in chunk (typically 2)
• —sentence_ms, sentence_en, etc.: Transcripts in each language

Usage

from datasets import load_dataset
import json

dataset = load_dataset("khursanirevo/multiturn_ks")

# Access audio and segments
chunk = dataset[0]
audio = chunk["audio"]  # Stereo audio array
video_id = chunk["video_id"]  # YouTube video ID
segments_en = json.loads(chunk["segments_en"])  # English speaker turns
segments_ms = json.loads(chunk["segments_ms"])  # Malay speaker turns

print(f"From video: {video_id}")
for seg in segments_en:
    speaker = seg['speaker']
    text = seg['text']
    print(f"Speaker {speaker} (EN): {text}")

for seg in segments_ms:
    speaker = seg['speaker']
    text = seg['text']
    print(f"Speaker {speaker} (MS): {text}")

Audio Format

• —Format: WAV (PCM)
• —Sample rate: 24kHz
• —Channels: 2 (stereo, speaker separation)
• —Bit depth: 32-bit float

Speaker Detection

Speakers are detected using RMS energy analysis:

• —Channel 0 (left): Speaker 0
• —Channel 1 (right): Speaker 1

Languages

Supported languages:

• —English (en)
• —Malay (ms)
• —Chinese Simplified (zh-Hans)
• —Chinese Traditional (zh-Hant)
• —Russian (ru)
• —Indonesian (id)
• —Arabic (ar)
• —Japanese (ja)
• —Korean (ko)

Dataset Statistics

• —Total videos: 139
• —Total chunks: 26383
• —Max chunk duration: 30s
• —Overlap: 0.5s

Source

Created from YouTube videos with dialogue separation using DialogueSidon model.

License

CC-BY-4.0