Automated system for generating minutes for Microsoft Teams meetings using multimedia processing and artificial intelligence

Identifier:  TFG:9472

Handle: https://hdl.handle.net/20.500.11797/TFG9472

Authors:  Ouladali, Assmaa

Abstract:
This paper presents an automated system for generating Microsoft Teams meeting minutes, addressing the inefficiencies and errors of manual documentation. Based on microservices and directed graphs in DOCs, it integrates downloading recordings from SharePoint, multimedia processing (evaluating FFmpeg vs. AWS), transcription and diary (Amazon Transcribe), analysis of video content using computer vision and LLMs, generation and correction of minutes (Claude via Bedrock), sending the minutes as PDFs (Outlook), and integration with the AIDA assistant. The results show structured minutes, the superiority of FFmpeg over AWS, and satisfactory accuracy in transcription and diary.