A comparative study of deep End-to-End Automatic Speech Recognition models for doctor-patient conversations in Polish in a real-life acoustic environment

Authors

  • Karolina Pondel-Sycz Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Marcin Lewandowski Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Jacek Olejnik JAS Technologie Sp. z o.o, Poland
  • Piotr Bilski Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Adrian Bilski Faculty of Applied Informatics and Mathematics, Warsaw University of Life Sciences, Poland https://orcid.org/0000-0003-4004-5484
  • Piotr Bobiński Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Leszek Morzyński Central Institute For Labour Protection-National Research Institute, Poland https://orcid.org/0000-0003-3534-3284
  • Emil Kozłowski Central Institute For Labour Protection-National Research Institute, Poland
  • Grzegorz Makarewicz Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Grzegorz Szczepański Central Institute For Labour Protection-National Research Institute, Poland https://orcid.org/0000-0003-0390-1624
  • Paweł Mazurek Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Maciej Jasiński Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Andrzej Buchowicz Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Agnieszka Paula Pietrzak Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
  • Iwona Olejnik JAS Technologie Sp. z o.o, Poland

DOI:

https://doi.org/10.24425/ijet.2025.153609

Abstract

The following paper presents research on the Automatic Speech Recognition (ASR) methods for the construction of a system to automatically transcribe the medical interview in Polish language during a visit in the clinic. Performance of four ASR models based on Deep Neural Networks (DNN) was evaluated. The applied structures included XLSR-53 large, Quartznet15x5, FastConformer Hybrid Transducer-CTC and Whisper large. The study was conducted on a self-developed speech dataset. Models were evaluated using Word Error Rate (WER), Character Error Rate (CER), Match Error Rate (MER), Word Accuracy (WAcc), Word Information Preserved (WIP), Word Information Lost (WIL), Levenshtein distance, Jaro - Winkler similarity and Jaccard index. The results show that the Whisper model outperformed other tested solutions in the vast majority of the conducted tests. Whisper achieved a WER = 20.84%, where XLSR-53 WER = 67.96%, Quartznet15x5 WER = 76.25%, FastConformer WER = 46.30%. These results show that Whisper needs further adaptation for medical conversations, as current volume of transcription errors is not practically acceptable (too many mistakes in the description of the patient's health description).

Published

2025-07-11

How to Cite

Pondel-Sycz, Karolina, et al. “A Comparative Study of Deep End-to-End Automatic Speech Recognition Models for Doctor-Patient Conversations in Polish in a Real-Life Acoustic Environment”. International Journal of Electronics and Telecommunications, vol. 71, no. 3, July 2025, doi:10.24425/ijet.2025.153609.

Issue

Section

Artykuły