A Unified Approach to Multilingual Automatic Speech Recognition with Improved Language Identification for Indic Languages

[Conference] INTERSPEECH 2024, Kos Island, Greece, September 2024

Authors:

Nikhil Jakhar, Sudhanshu Srivastava, Arun Baby

Abstract:

This work presents an integrated framework combining language identification with multilingual ASR using the Whisper architecture. The proposed approach achieves an absolute 19.1% improvement in Word Error Rate (WER) while enhancing language identification performance by 6% in terms of Diarization Error Rate (DER) for Indic languages.

Cite:

@inproceedings{jakhar24_interspeech,
  title = {{A Unified Approach to Multilingual Automatic Speech Recognition with Improved Language Identification for Indic Languages}},
  author = {Nikhil Jakhar and Sudhanshu Srivastava and Arun Baby},
  year = {2024},
  booktitle = {{Interspeech 2024}},
  pages = {3949--3953},
  doi = {10.21437/Interspeech.2024-2043}
}

Links:

Proceedings

PDF