Whisper-small-ta Open-source Speech Recognition Model - Free Deployment for Precise Tamil Speech Recognition

Whisper Small Ta

Developed by navin-kumar-j

This model is a speech recognition model fine-tuned on the Tamil Common Voice 17.0 dataset based on OpenAI's Whisper Small, with a Word Error Rate (WER) of 43.23%.

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Tamil speech recognition #Low-resource speech processing #Whisper fine-tuning

Downloads 38

Release Time : 3/27/2025

Model Overview

This is an automatic speech recognition (ASR) model optimized for Tamil, fine-tuned based on the Whisper Small architecture, suitable for Tamil speech-to-text tasks.

Model Features

Tamil Optimization

Specifically fine-tuned for Tamil, offering better performance in Tamil recognition compared to the original Whisper model

Medium Scale

Based on the Whisper Small architecture, striking a balance between computational resources and performance

Open Source License

Uses Apache 2.0 license, allowing for both commercial and research use

Model Capabilities

Tamil speech recognition

Speech-to-text

Audio content transcription

Use Cases

Speech Transcription

Tamil Meeting Minutes

Automatically transcribe Tamil meeting recordings into text records

Word Error Rate approximately 43.23%

Voice Assistant

Used as a speech recognition component for Tamil voice assistant applications

Education

Language Learning Assistance

Help Tamil learners check pronunciation and speech accuracy

🚀 Whisper Small Ta - Navin Kumar J

This is a fine - tuned model based on the openai/whisper - small architecture, trained on the Common Voice 17.0 dataset, aiming to provide high - quality automatic speech recognition for Tamil.

🚀 Quick Start

This model is a fine - tuned version of [openai/whisper - small](https://huggingface.co/openai/whisper - small) on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

Loss: 0.2155
Wer: 43.2294

📚 Documentation

Model Information

Property	Details
Library Name	transformers
Language	ta
License	apache - 2.0
Base Model	openai/whisper - small
Tags	generated_from_trainer
Datasets	mozilla - foundation/common_voice_17_0
Metrics	wer

Model Index

Name: Whisper Small Ta - Navin Kumar J
Results:
- Task:
  - Name: Automatic Speech Recognition
  - Type: automatic - speech - recognition
- Dataset:
  - Name: Common Voice 17.0
  - Type: mozilla - foundation/common_voice_17_0
  - Config: ta
  - Split: None
  - Args: 'config: ta, split: test'
- Metrics:
  - Name: Wer
  - Type: wer
  - Value: 43.22944067339895

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e - 05
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon = 1e - 08 and optimizer_args = No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 4000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1639	0.2773	1000	0.2724	51.7611
0.1547	0.5546	2000	0.2353	46.9249
0.1357	0.8319	3000	0.2243	44.6897
0.0967	1.1093	4000	0.2155	43.2294

Framework versions

Transformers 4.50.1
Pytorch 2.6.0+cu124
Datasets 3.4.1
Tokenizers 0.21.1

📄 License

This project is licensed under the apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご