3 Speech to text phase – Errors categories and examples скачать в хорошем качестве

3 Speech to text phase – Errors categories and examples

Under the academic supervision of Ca’ Foscari University of Venice The video describes how an AI-based system automatically transcribes spoken Arabic and shows both its advantages and its limits. It examines the main error categories and analyses common transcription issues, such as punctuation, unclear pronunciation, foreign terminology, multiple speakers, background music, dialectal variations, and difficulties in recognizing proper names, grammatical structures, and direct speech. Overall, the video shows why human revision remains essential to ensure accurate and reliable transcriptions. “Artificial Intelligence for Multilingual Radio Broadcasting” is a project based on the radio series “Arab Philosophers – Ancient and Contemporary” a coproduction led by COPEAM and ASBU. The project is supported by UNESCO in collaboration with, Ca’ Foscari University/Venice and Rai Centre for Research, Technological Innovation and Experimentation (CRITS).