Automatic post-editing of subtitles - Rule-based post-editing of subtitles from Swiss German to Standard German

Diplôme Master Automatic post-editing of subtitles - Rule-based post-editing of subtitles from Swiss German to Standard German eng 2022 author Haberkorn Veronika Christine 920029 director Bouillon Pierrette 0000-0002-8854-6360 26962 This master thesis is part of the PASSAGE project, which aims to develop a system to automatically generate Standard German subtitles for spoken Swiss German in collaboration with the « Schweizer Radio and Fernsehen » (SRF). For this task, a normalized transcription from speech recognition needs to be post-edited for the training of an automatic post-editing system. The goal of the thesis is to develop and evaluate a rule-based automatic post-editing system, which is intended to assist post-editors in generating corrected segments. The system exploits regular expressions and rules using the natural language processing library spaCy. The developed rules cover errors due to spoken language and specific Swiss German phenomenons. The system detected errors (performed modifications) in approx. 5% (4%) of the normalized transcription input. A human evaluation confirmed the necessity (correctness) of the modifications in 97% (88%) of the cases. Rule-based post-editing Automatic post-editing APE Swiss German Regular expressions Regex SpaCy Natural language processing Nlp Dewey 410.2 Maîtrise universitaire en traitement informatique multilingue 112 Département de traitement informatique multilingue 92 false ac36cd84-2a6c-42a8-85d5-825be9f61419 Master thesis Public Veronika_Haberkorn_master_thesis.pdf 3409507 application/pdf openAccess Faculté de traduction et d'interprétation / Département de traitement informatique multilingue

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Automatic post-editing of subtitles - Rule-based post-editing of subtitles from Swiss German to Standard German				2022	345	419

Archive ouverte UNIGE

Haberkorn, Veronika Christine