JBT Journal
Advanced

Raqamli Transformatsiya va Sun’iy Intellekt

NUTQNI TANIB OLISHDA CHUQUR OʻQITISH MODELLARI TAHLILI

Primqulov Oybek Dilmurodovich1, Sherqoʻziyeva Shahnozabonu Baxromjon qizi1

1 Muhammad al-Xorazmiy nomidagi Toshkent axborot texnologiyalari universiteti

Zenodo DOI: 10.5281/zenodo.20590503

Keywords

transformer Whisper. kvantlash Self-Supervised Learning Deep Speech 2 avtomatik nutqni tanib olish Whisper Wav2Vec2.0 nutq texnologiyalari tashqi qurilmalar

Abstract

Ushbu maqolada avtomatik nutqni tanib olish sohasida qoʻllanilayotgan zamonaviy chuqur oʻqitish modellari tahlil qilingan. Tadqiqotda Self-Supervised Learning yondashuviga asoslangan Wav2Vec 2.0 modelining ishlash tamoyillari, arxitekturasi va amaliy imkoniyatlari koʻrib chiqilgan. Modelning kam hajmdagi annotatsiyalangan ma’lumotlar bilan yuqori aniqlikka erishish qobiliyati LibriSpeech natijalari asosida baholangan. Shuningdek, Wav2Vec 2.0 modelining Raspberry Pi kabi cheklangan resurslarga ega qurilmalardagi samaradorligi, kvantlashtirish usullarining hisoblash xarajatlariga ta’siri ham tahlil qilingan. Tadqiqot davomida Deep Speech 2 va Whisper modellari bilan qiyosiy tahlil oʻtkazilib, ularning afzallik va cheklovlari yoritilgan. Olingan natijalar Wav2Vec 2.0 modelining kam hajmdagi nutq ma’lumotlari bilan ham samarali ishlashini, hisoblash resurslariga boʻlgan talabni optimallashtirish imkoniyatiga ega ekanligini hamda turli amaliy sohalarda qoʻllash uchun istiqbolli yechim sifatida qaralishini koʻrsatadi.

Views: 173 Published: 2026-06-25 Pages: 80-84 Volume: 4 Issue: 3

References

  1. Dario Amodei at all. Deep Speech 2: End-to-End Speech Recognition in English and Mandarin. https://arxiv.org/abs/1512.02595v1 8 Dec 2015.
  2. Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli. wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations. https://arxiv.org/abs/2006.11477v3
  3. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever. Robust Speech Recognition via Large-Scale Weak Supervision. https://arxiv.org/abs/2212.04356v1 6 Dec 2022
  4. Santosh Gondi, Wav2Vec2.0 on the Edge: Performance Evaluation. Meta Inc. arXiv:2202.05993v1 [cs.SD] 12 Feb 2022. Sound (cs.SD); Human-Computer Interaction(cs.HC); Audio and Speech Processing (eess.AS) https://doi.org/10.48550/arXiv.2202.05993
  5. Juan Camilo Vásquez-Correa, Aitor Álvarez Muniain. Novel Speech Recognition Systems Applied to Forensics within Child Exploitation: Wav2vec2.0 vs. Whisper. Journal of Sensors, Vol.23, Issue 4, 2023. https://doi.org/10.3390/s23041843
  6. Vassil Panayotov, Guoguo Chen, Daniel Povey, S. Khudanpur. LibriSpeech: An ASR corpus based on public domain audio books. Journal of IEEE International Conference on Acoustics, Speech, and Signal Processing.