Master
OA Policy
English

(Automatic) text simplification of subtitles: Research Questions: Is it possible to make functioning Easy Language subtitles? Can the machine also simplify the subtitles to a certain extent? Will the target group prefer the human or the machine translation?

ContributorsArnold, Melanie
Master program titleMaîtrise universitaire en traduction et technologies
Defense date2024
Abstract

This thesis researches the impact of subtitle simplification on comprehension and audience satisfaction across two different target groups of Easy Language. It compares three subtitle simplification approaches: Manual simplification, limited simplification with the Passage system, and ChatGPT-based simplifications. It assesses the feasibility of Easy Language for subtitles and the levels of comprehension by the language group. The study begins with an in-depth analysis of Easy Language in German and establishes customized guidelines for future simplification efforts. Utilizing the Passage project's curated corpus, three text extracts are simplified using manual methods, the Passage tool, and ChatGPT. These simplified versions were then evaluated by both target audiences and industry experts. This thesis found that adapting subtitles into Easy Language is feasible and particularly effective for individuals with Alzheimer’s. Manual simplification proved to be the most efficient method, however ChatGPT was also well suited for the simplification in most cases. All the methods require finetuning to address specific audience needs.

Keywords
  • Easy Language
  • Text simplification
  • Automatic Text Simplification
  • ChatGPT
  • Passage system
  • Manual Simplification
  • Controlled Language
Citation (ISO format)
ARNOLD, Melanie. (Automatic) text simplification of subtitles: Research Questions: Is it possible to make functioning Easy Language subtitles? Can the machine also simplify the subtitles to a certain extent? Will the target group prefer the human or the machine translation? Master, 2024.
Main files (1)
Master thesis
accessLevelPublic
Identifiers
  • PID : unige:179122
59views
12downloads

Technical informations

Creation17/07/2024 12:43:08
First validation02/08/2024 08:53:42
Update time18/10/2024 06:20:38
Status update18/10/2024 06:20:38
Last indexation01/11/2024 10:40:47
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack