W

Wav2vec2 Xls R 300m Mr Cv9 With Lm

Developed by anuragshas
An automatic speech recognition model fine-tuned on Marathi speech datasets based on Facebook's XLS-R-300M model
Downloads 23
Release Time : 5/8/2022

Model Overview

This model is an optimized automatic speech recognition (ASR) system for Marathi, fine-tuned on the Common Voice 9.0 Marathi dataset, capable of converting Marathi speech into text.

Model Features

High-accuracy Marathi recognition
Achieves a word error rate of 23.841% and a character error rate of 5.522% on the Common Voice 9.0 test set
Based on large-scale pre-trained model
Fine-tuned from the XLS-R-300M model, inheriting powerful speech feature extraction capabilities
Optimized training process
Uses techniques like linear learning rate scheduling and gradient accumulation to optimize training results

Model Capabilities

Marathi speech recognition
Speech-to-text
Audio content transcription

Use Cases

Speech transcription
Marathi speech transcription
Convert Marathi speech content into text format
Achieves a word error rate of 23.841% on the test set
Voice assistants
Marathi voice command recognition
Used to develop voice assistants and interactive systems supporting Marathi
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase