Analyse de la qualité du sous-titres automatiques sur la plateforme YouTube

Master program titleMaîtrise universitaire en traduction et technologies (MATT)
Defense date2023

In the digital landscape, where multimedia content prevails, subtitling and speech recognition play vital roles in facilitating effective communication. This study investigates the perceptual contrast between human-crafted subtitles and those generated via speech recognition on YouTube. With a hypothesis favoring human-authored subtitles, the research examines 30 subtitle lines from videos, comparing manual and automatic subtitling. Evaluation entails WER (Word Error Rate) and TER (Translation Edit Rate) scores for automated assessment and human evaluators utilizing the NER (Number, Edition and Recognition Error) score. Findings emphasize the necessity of robust technological advancements to enhance user experiences and ensure broad accessibility to diverse content.

  • YouTube
  • Speech recognition
  • Subtitling
  • Word Error Rate
  • Translation Edit Rate
  • Number Edition and Recognition Error
Citation (ISO format)
MORALES DE LA TORRE, Juan Mario. Analyse de la qualité du sous-titres automatiques sur la plateforme YouTube. 2023.
Main files (1)
Master thesis
  • PID : unige:174145

Technical informations

Creation01/11/2024 1:45:06 PM
First validation01/12/2024 7:21:29 AM
Update time03/28/2024 7:26:07 AM
Status update03/28/2024 7:26:07 AM
Last indexation05/06/2024 5:44:19 PM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack