MDT-AE067 Korean Conversational Speech Corpus
*
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Language
Korean
Recording Environment
Indoor
Duration
1,000 h
Data Content
Themed conversations
File Format
WAV, TXT
Speech Parameter
16 kHz, 16 bits
Recording Equipment
Mobile
Field of Application
ASR
Data Sensitive Items
nil
Copyright Owner
Magic Data
open source duration
5.22 h
Sample
No related data? Contact us
Related Datasets
MDT-AE067 Korean Duplex Conversation Training Dataset
Leveraging dual-track recording technology to accurately separate overlapping speech and spontaneous interruptions, this dataset authentically reflects the unique honorific hierarchy, emotional sentence endings, and rapid turn-taking found in Korean. It includes multi-speaker role annotations and scenario tags, enabling AI to better understand the intricate conversational dynamics and emotional subtleties embedded in Korean cultural contexts.
MDT-AD006 Mandarin Chinese Conversational Speech Corpus
Play Audio
MDT-AD015 English Conversational Telephone Speech Corpus—Telephony
[Open-Source]
Play Audio
MDT-AF027 Brazilian Portuguese Conversational Speech Corpus
Play Audio
Multi-Turn Conversation Dataset for Speech Language Model
Play Audio
MDT-RI001 Chinese Spoken Speech Dataset
This dataset is designed to train AI models for better spoken language understanding, enhancing natural interaction in Chinese speech recognition. It features real-world conversations across diverse scenarios, recorded by a wide range of speakers, with high transcription accuracy. All utterances retain full prosodic characteristics of spoken Chinese, with detailed pause and punctuation annotations to help models learn natural rhythm and improve interaction fluency.
Why MD Datasets
Full Compliance
ISO/IEC 27001 & ISO/IEC 27701:2019 compliant
Multiple Dimension
Audio, text, image, and video multi-modal data
Extensive Scope
Conversational, scripted, and spontaneous data covering extensive domains
High Accuracy
Expertise secured quality result