img Prof. Samir Rustamov
ADA University, School of IT & Engineering (SITE), Baku, Azerbaijan
Head of AI Laboratory, MegaSec Company, Baku, Azerbaijan

Building Speech Recognition Systems that Work in the Real World
Recent advances in deep learning and large-scale data have significantly improved speech recognition accuracy on benchmarks, yet deploying reliable systems in real-world environments remains challenging. This keynote focuses on building speech recognition systems that function effectively outside controlled settings, addressing issues such as noise, accents, domain mismatch, and real-time constraints. It examines the end-to-end speech recognition pipeline, highlights practical failure modes, and questions the adequacy of traditional evaluation metrics like word error rate. The talk also explores the integration of speech recognition with broader AI systems, including natural language understanding and large language models, while emphasizing ethical, social, and usability considerations. The key message is that real-world speech recognition is a system-level challenge that extends beyond model performance to include data quality, robustness, and user trust.

Speakers

Speakers

  • img
    Prof. Boris Mordukhovich
    Institute for AI and Data Science, Wayne State University (USA)

    Abstract >>>
  • img
    Prof. Sedat Akleylek
    Institute of Computer Science, University of Tartu (Estonia)

    Abstract >>>
  • img
    Prof. Perviz Ahmedzade
    Ege University (Türkiye),
    Samarkand State Architectural and Civil Engineering University (Uzbekistan)
    Abstract >>>

  • img
    Prof. Azer Kasimzade
    Azerbaijan University of Architecture and Construction (Azerbaijan)
    Abstract >>>
  • img
    Prof. Semyon Serovaysky
    Al-Farabi Kazakh National University (Kazakhstan)
    Abstract >>>
  • img
    Prof. Samir Rustamov
    ADA University (Azerbaijan)
    Abstract >>>