-
Model Overview:
This is a fine-tuned version of OpenAI's Whisper model, trained to perform speech-to-text translation from Indic languages to English. The model has been optimized using open-source datasets and proprietary dataset created by AbleCredit.com to enhance transcription quality for Indic languages.
-
Languages Supported
The model is trained to handle speech-to-text translation for the following Indic languages:
- Hindi (hi)
- Marathi (mr)
- Gujarati (gu)
- Tamil (ta)
- Telugu (te)
- Kannada (kn)
The base Whisper model was fine-tuned using a carefully curated dataset consisting of open-source corpus and a well curated internal dataset from AbleCredit. The dataset underwent preprocessing, including text normalization, score based filtering and feature engineering.
-
Key Improvements After Fine-Tuning
📉 Word Error Rate (WER) Reduction: 54% lower WER compared to the base model.
📈 BLEU Score Increase: 5.5× higher BLEU score, indicating improved translation accuracy.
-
Word Error Rate (WER)
WER is used to measure the error rate in transcriptions. A lower WER means better transcription accuracy.
WER Reduction: 54% improvement over the base Whisper model.
-
BLEU Score
The BLEU (Bilingual Evaluation Understudy) Score is a metric used to evaluate machine translation quality. Higher BLEU scores indicate better translation quality.
BLEU Score Improvement: 5.5× higher BLEU compared to the base model.
