Ez minutes: Minutes of the meeting transcriber/ Elizabeth M. Guevarra, Bryan Aiman O. Manalaysay, Alexander G. Rosete, and Kyle Vencel P. So.--

Bachelor's thesis

College of Science.-- Bachelor of science in computer science: Technological University of the Philippines, 2025.

Includes bibliographic references and index.

This study aimed to develop a system that automatically extracts and generates a
structured meeting summary from recorded meeting sessions in both audio and video
format. The primary goal of the application is to streamline the process of meeting
documentation by automating transcription, agenda/topic identification, and
summarization into a well-organized minutes document. This research supported both
English and Filipino languages, and code-switched combinations of both that will prove to
be a big help for meetings held academically and professionally in the Philippines. The
system made use of the following tools to ensure it carries out its goals and objectives:
FFmpeg for audio extraction, Faster Whisper for speech-to-text transcriptions, Pyannote
for speaker diarization, FastText for language identification, a Language Model (LM) by
META, Llama 3.1 8b, for agenda detection with other principles for LM and NLP to make
the generation of meeting minutes possible, feasible, and accurate. The system was
evaluated using the ISO 25010 criteria in terms of the system's functional stability,
performance efficiency, usability, reliability, and maintainability. The system was graded
with an acceptability rating of 3.42 that is interpreted as "Highly Acceptable".