AIbase
EN
Home
Timit 5percent Supervised
T

Timit 5percent Supervised

Developed by Kuray107
A speech recognition model fine-tuned on the TIMIT dataset based on facebook/wav2vec2-large-lv60, using 5% of the data for supervised training
Speech Recognition
Transformers
Open Source License:Apache-2.0#Speech Recognition#Low-resource Fine-tuning#Wav2Vec2 Architecture
Downloads 31
Release Time : 3/2/2022

Model Overview

This model is a speech recognition model specifically optimized for English speech recognition tasks, achieving a low word error rate on the TIMIT dataset.

Model Features

Efficient Supervised Learning
Achieves good recognition performance using only 5% of the TIMIT dataset for training
Low Word Error Rate
Achieves a word error rate of 27.88% on the evaluation set
Based on wav2vec2 Architecture
Uses facebook's wav2vec2-large-lv60 as the base model, featuring powerful speech feature extraction capabilities

Model Capabilities

English Speech Recognition
Speech-to-Text
Continuous Speech Recognition

Use Cases

Speech Transcription
Meeting Minutes Transcription
Automatically transcribe English meeting recordings into text records
Achieves a 27.88% word error rate on the TIMIT test set
Voice Command Recognition
Recognize English voice commands
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase