wav2vec2-large-xlsr-53-telugu
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Author | anuragshas |
| Test WER | 44.98% |
| Dataset | OpenSLR SLR66 |
What is wav2vec2-large-xlsr-53-telugu?
This model is a fine-tuned version of Facebook's wav2vec2-large-xlsr-53 specifically optimized for Telugu speech recognition. It leverages the OpenSLR SLR66 dataset and is designed to process audio inputs sampled at 16kHz. The model demonstrates practical application in automatic speech recognition (ASR) for Telugu language processing.
Implementation Details
The model is implemented using PyTorch and the Transformers library, utilizing the wav2vec2 architecture. It was trained on 70% of the OpenSLR Telugu dataset and achieved a Word Error Rate (WER) of 44.98% on the test set. The implementation includes built-in audio resampling capabilities and preprocessing functions for handling Telugu text normalization.
- Supports 16kHz audio input processing
- Includes custom text normalization for Telugu
- Implements batch processing for efficient inference
- Provides direct integration with the Transformers library
Core Capabilities
- Automatic Speech Recognition for Telugu language
- Real-time audio processing and transcription
- Batch processing support for multiple audio files
- Custom text normalization and preprocessing
Frequently Asked Questions
Q: What makes this model unique?
This model is specifically optimized for Telugu language speech recognition, building upon the powerful wav2vec2-large-xlsr-53 architecture. It includes custom preprocessing and normalization specific to Telugu text processing.
Q: What are the recommended use cases?
The model is ideal for Telugu speech recognition tasks, particularly in applications requiring 16kHz audio input. It's suitable for transcription services, voice assistants, and other Telugu language processing applications.
