W

Whisper Small Ko

Developed by SungBeom
Korean speech recognition model based on the Whisper Small architecture, fine-tuned on multi-domain Korean datasets
Downloads 524
Release Time : 6/19/2023

Model Overview

This model specializes in Korean automatic speech recognition tasks, trained on large-scale multi-domain datasets with excellent domain adaptability

Model Features

Multi-domain Joint Training
Simultaneously trained on datasets from multiple domains including customer service calls, in-car conversations, and welfare services to enhance cross-domain recognition capabilities
Large-scale Data Training
Trained on over 10,916,423 samples (approximately 13,946 hours), covering various speech scenarios
Adapter Technology Compatibility
Supports further improvement in dialect, children's, and elderly speech recognition through adapter technology

Model Capabilities

Korean speech-to-text
Multi-domain speech recognition
Customer service scenario speech recognition
In-car scenario speech recognition

Use Cases

Customer Service Scenarios
Customer Call Transcription
Automatically transcribes customer service call content
Word error rate 9.48%
In-car Systems
In-car Voice Command Recognition
Recognizes voice commands in in-car environments
Social Welfare
Welfare Consultation Recording
Automatically records welfare consultation call content
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase