AIbase
EN
Home
84rry Xls R 300M AR
8

84rry Xls R 300M AR

Developed by 84rry
This model is a fine-tuned Arabic speech recognition model based on facebook/wav2vec2-xls-r-300m on the Common Voice dataset, achieving a word error rate of 0.5078 on the evaluation set.
Speech Recognition
Transformers
Open Source License:Apache-2.0#Arabic speech recognition#Large model fine-tuning#Low word error rate
Downloads 27
Release Time : 6/9/2022

Model Overview

This is a fine-tuned model for Arabic speech recognition, based on Facebook's wav2vec2-xls-r-300m architecture and trained on the Common Voice dataset.

Model Features

Based on XLS-R architecture
Utilizes Facebook's large-scale cross-lingual speech representation learning architecture XLS-R, which has powerful speech feature extraction capabilities.
Arabic language optimization
Specially fine-tuned and optimized for Arabic speech, performing well on the Common Voice dataset.
Medium-sized model
300M parameter scale, striking a balance between performance and computational resource requirements.

Model Capabilities

Arabic speech recognition
Speech-to-text
Audio content transcription

Use Cases

Speech transcription
Arabic speech-to-text
Convert Arabic speech content into text format
Word error rate 0.5078
Voice assistant
Arabic voice command recognition
Used for Arabic voice assistants or control systems' voice command recognition
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