This model is a fine-tuned Automatic Speech Recognition (ASR) model developed for recognizing the Rayalaseema dialect of Telugu. It is based on the Wav2Vec2 architecture and fine-tuned on Telugu speech collected through the Swecha Gonthuka initiative.
Model Details
Task: Automatic Speech Recognition (ASR)
Architecture: Wav2Vec2
Framework: Hugging Face Transformers
Language: Telugu (te)
Dialect: Rayalaseema
Base Model: swechatelangana/swecha-gonthuka-asr
Intended Use
This model is intended for:
Telugu speech transcription
Rayalaseema dialect recognition
Speech dataset research
Educational and accessibility applications
Training Data
The model was fine-tuned using Telugu speech recordings collected from native speakers of the Rayalaseema dialect.
Evaluation
The model was evaluated using Word Error Rate (WER). Performance depends on audio quality and speaker characteristics.
Limitations
Designed primarily for Rayalaseema Telugu.
Performance may decrease for other Telugu dialects.
Background noise can reduce transcription accuracy.
Citation
If you use this model in your work, please cite the Swecha Gonthuka project.
