A Zero-Shot Dual-LLM Pipeline for End-to-End Indonesian Meeting Summarization: A Benchmark Study

DOI

10.2991/978-94-6463-926-1_74How to use a DOI?

Keywords

Indonesian Automatic Speech Recognition; Large Language Models; Meeting Summarization

Abstract

Meetings are essential for organizational decision-making, yet manual minute-Meetings play a critical role in organizational decision-making; however, manual minute-taking often suffers from inaccuracy and inefficiency. This study introduces Smart Notulensi, a zero-shot dual-LLM pipeline for end-to-end Indonesian meeting summarization in low-resource settings, addressing challenges of data scarcity and linguistic diversity in low-resource Indonesian NLP. We benchmark three state-of-the-art ASR models (Gemini 1.5 Pro, Deepgram Nova-2, and Azure Speech) on real-world multi-speaker Indonesian meeting, an evaluate transcription accuracy with Word Error Rate (WER) and speaker attribution with Speaker Attribution Error Rate (SAER). Gemini 1.5 Pro achieves the best performance with average WER and SAER of 9.85% and 6.21%, respectively. The resulting transcripts are summarized using Claude 3.5 Sonnet, yielding structured minutes with ROUGE-1, ROUGE-2, and ROUGE-L F1-scores of 0.848, 0.677, and 0.806. Implemented in a Streamlit-based web application, the proposed pipeline demonstrates robust performance across diverse meeting scenarios and establishes a strong baseline for automated meeting documentation in low-resource languages. Future work will focus on improving dialogue complexity handling and expanding dataset diversity.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Tita Karlita
AU  - Ronny Susetyoko
AU  - Bangkit Muhammad Najib
PY  - 2025
DA  - 2025/12/31
TI  - A Zero-Shot Dual-LLM Pipeline for End-to-End Indonesian Meeting Summarization: A Benchmark Study
BT  - Proceedings of the International Conference on Applied Science and Technology on Engineering Science 2025 (iCAST-ES 2025)
PB  - Atlantis Press
SP  - 658
EP  - 667
SN  - 2352-5401
UR  - https://doi.org/10.2991/978-94-6463-926-1_74
DO  - 10.2991/978-94-6463-926-1_74
ID  - Karlita2025
ER  -