Speech-driven Abstractive Summarization For Automated Meeting Minutes Generation

Abstract

Meetings Are An Essential Part Of The Decision-making Process Of An Organization. However, Manually Writing The Minutes Of Meeting (MoM) Is A Laborious And Time-consuming Task. This Paper Proposes An Automated System For Generating The Minutes Of Meeting From The Meeting Audio. The System Uses Automatic Speech Recognition (ASR) And Natural Language Processing (NLP) For Generating A Well-structured And Concise Textual Summary Of The Meeting. The System Uses The Open AI Whisper ASR Model To Generate The Text From The Multi-speaker Meeting Audio. After Text Generation, The Text Is Processed, And Then The BERT-based Text Summarization Model Is Used To Generate A Well-structured MoM. The Results Of The Experiment Show That The Proposed System Can Generate The MoM Document Very Quickly, With Acceptable Accuracy. The System Can Be Used For Real-time And Offline Applications, Which Can Be Used For Academic And Corporate Environments.

Keywords