S

Sew D Tiny 100k Ft Ls100h

Developed by asapp
SEW-D-tiny is an efficient speech recognition pre-trained model developed by ASAPP Research, focusing on the balance between performance and efficiency.
Downloads 24.55k
Release Time : 3/2/2022

Model Overview

This model is pre-trained on 16kHz sampled speech audio and is suitable for downstream tasks such as automatic speech recognition, speaker recognition, and intent classification.

Model Features

Efficient Inference
Achieves 1.9x inference speedup compared to wav2vec 2.0.
Performance Improvement
Reduces word error rate by 13.5% relative in the semi-supervised setting of LibriSpeech 100h-960h.
Lightweight
The model design emphasizes efficiency, making it suitable for resource-constrained environments.

Model Capabilities

Speech Recognition
Speaker Recognition
Intent Classification
Emotion Recognition

Use Cases

Speech-to-Text
LibriSpeech Transcription
Convert speech in the LibriSpeech dataset to text.
Achieves a WER of 10.47 on the LibriSpeech clean test set and 22.73 on the other test set.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase