Smart Technology and Artificial Intelligence
Multilingual Meeting Transcription and Summarization Using Machine Learning
Authors: Nihar Shetty S, M J Akshaya, Ankitaa Pupneja and Siddharth Kumar
Publishing Date: 04-06-2026
ISBN: 978-81-975670-2-5
Abstract
The need for accurate documentation of multilingual discourse has been steadily increasing in the global business and academic community. However, the spoken modality is often limited by code-switching, which is the seamless blending of English with local languages such as Hindi, Kannada, and Punjabi, making it difficult to process in the traditional monolingual fashion. While highly advanced deep learning architectures for speech recognition are available, their weakness in maintaining semantic consistency in code-switched settings often hinders their usability for meeting summarization tasks. This paper proposes a comprehensive AI-based solution to overcome the linguistic performance and efficiency trade-off by leveraging the power of OpenAI’s Whisper for high-quality transcription and Meta’s Llama 3 for abstractive summarization. The proposed solution optimizes transcription for Indic code-switched speech and uses a generative model to generate summaries in the user-preferred regional scripts. A comprehensive performance evaluation is carried out, and the proposed system is compared with the IIT Bombay Code-Switching Dataset and the traditional extractive summarization baselines. The quantitative evaluation shows a significant improvement in the Word Error Rate (WER) and a significant improvement in ROUGE-L coherence scores, competitive summarization performance, while still being capable of processing and generating summaries in multiple regional languages. The proposed system is shown to be a viable option for inclusive professional settings, providing a realistic balance between linguistic performance and cross-platform usability for real-world multilingual meeting summarization tasks
Keywords
Multilingual Transcription, Abstractive Summarization, Whisper ASR, Code-Switching, Indic Languages, Llama 3, Natural Language Processing.
Cite as
Nihar Shetty S, M J Akshaya, Ankitaa Pupneja and Siddharth Kumar, "Multilingual Meeting Transcription and Summarization Using Machine Learning", In: Mukesh Saraswat, Sandeep Kumar, Manjunatha Sughaturu Krishnappa and Rakesh Keshava (eds), Smart Technology and Artificial Intelligence, SCRS, India, 2026, pp. 53-63. https://doi.org/10.56155/978-81-975670-2-5-5