B

Bp Voxforge1 Xlsr

Developed by lgris
This is a Wav2Vec2 model fine-tuned for Brazilian Portuguese speech recognition tasks, trained on the VoxForge dataset.
Downloads 21
Release Time : 3/2/2022

Model Overview

This model is based on Facebook's Wav2Vec2 architecture, specifically fine-tuned for Brazilian Portuguese speech recognition tasks. It can convert Portuguese speech into text and is suitable for various Brazilian Portuguese dialects.

Model Features

Multi-dataset evaluation
The model has been comprehensively evaluated on multiple Brazilian Portuguese datasets, including CETUC, Common Voice, and 7 other different datasets.
Language model integration
Supports integration with a 4-gram language model, significantly reducing the word error rate (WER).
Lightweight solution
Trained on a relatively small VoxForge dataset (3.9 hours) but still achieves decent recognition performance.

Model Capabilities

Brazilian Portuguese speech recognition
Speech-to-text
Supports various Brazilian dialects

Use Cases

Speech transcription
Brazilian Portuguese speech transcription
Convert Brazilian Portuguese speech content into text
Average word error rate of 0.584 (without language model) or 0.454 (with 4-gram language model)
Voice assistants
Brazilian Portuguese voice command recognition
Basic recognition component for Brazilian Portuguese voice assistants
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase