An end-to-end system for transcription, translation, and summarization to support the co-creation process. A Health CASCADE Study.
George Balaskas, Homer Papadopoulos, Quentin Emile Armand Loisel, Dimitra Pappa, George P. Efthymoglou, SEBASTIEN F. M. CHASTIN · 2023
This paper presents a web service and a deep learning (DL) pipeline that has been developed and implemented as part of the MSCA Health CASCADE project. The purpose of the web service is to provide support, streamline, and enable participatory methods, such as co-creation, in the public health domain. The DL pipeline assists with translation, transcription, speaker diarization, relation extraction, and summarization of audio recordings. This is achieved by removing the need for time-consuming tasks, such as translating and transcribing audio recordings. Additional value is created by extracting implicit relations from the transcribed text and identifying patterns, key themes, and trends. Finally, providing summaries of the transcripts creates a sense of ownership that can improve stakeholder retention in participatory methods.