EUVIMEDEUVIMED entdecken
Crossref

Optimizing ASR Models through Contrastive Learning-Based Audio Filtering

Pranay. S Awachar, Prof. Shivani.V Khedkar, Rushikesh. G Kothare, Rohit. S Ambhore, Nitin. R Diware

International Scientific Journal of Engineering and Management · 19.04.2026 · 05 (04) · 1-6

DOI: 10.55041/isjem06542 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Scientific Journal of Engineering and Management, online erschienen am 19.04.2026. Verfasst von Pranay. S Awachar und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a novel methodology for enhancing Automatic Speech Recognition (ASR) performance by utilizing contrastive learning to filter synthetic audio data. We address the challenge of incorporating synthetic data into ASR training, especially in scenarios with limited real- world data or unique linguistic characteristics. The method utilizes a contrastive learning model to align representations of synthetic audio and its corresponding text transcripts, enabling the identification and removal of low-quality samples that do not align well semantically. We evaluate the methodology on a medium-resource language across two distinct datasets: a general-domain dataset and a regionally specific dataset characterized by unique pronunciation patterns. Experimental results reveal that the optimal filtering strategy depends on both model capacity and dataset characteristics. Larger models, like Whisper Large V3, particularly benefit from aggressive filtering, while smaller models may not require such stringent filtering, especially on non- normalized text. This work highlights the importance of adjusting synthetic data augmentation and filtering to specific model architectures and target domains. The proposed method, robust and adaptable, enhances ASR performance across diverse language settings. We have open-sourced the entire work, which includes 140 hours of synthetically generated Portuguese speech, as well as the pipeline and parameter settings used to create these samples. Additionally, we provide the fine-tuned Whisper models and the code required to reproduce this research. ai/semantic_audio_filtering.

Zitieren

APA

Awachar, P. S., Khedkar, P. S., Kothare, R. G., Ambhore, R. S., & Diware, N. R. (2026). Optimizing ASR Models through Contrastive Learning-Based Audio Filtering. International Scientific Journal of Engineering and Management, 05(04), 1-6. https://doi.org/10.55041/isjem06542

Vancouver

Awachar PS, Khedkar PS, Kothare RG, Ambhore RS, Diware NR. Optimizing ASR Models through Contrastive Learning-Based Audio Filtering. International Scientific Journal of Engineering and Management. 2026;05(04):1-6. doi:10.55041/isjem06542

BibTeX
@article{Awachar2026,
  author = {Awachar, Pranay. S and Khedkar, Prof. Shivani.V and Kothare, Rushikesh. G and Ambhore, Rohit. S and Diware, Nitin. R},
  title = {Optimizing ASR Models through Contrastive Learning-Based Audio Filtering},
  journal = {International Scientific Journal of Engineering and Management},
  year = {2026},
  volume = {05},
  number = {04},
  pages = {1--6},
  doi = {10.55041/isjem06542}
}
Kostenlos starten →