Audio Transcription of Meetings and Action Items using LLM’s

Large Language Models (LLMs) have garnered significant attention as a state-of-art tool for textual data processing and analysis, owing to their unparalleled capacity for generating and comprehending text at a degree of sophistication never seen before [1]. In the realm of meeting management, there are numerous instances of overseeing structured derivatives, leading to missed opportunities for appropriate job delegation and follow-up. This paper describes the use of llama-2 model and technologies like voice transcription, text processing, and action item recognition approaches to streamline the process of audio transcription and precise action item generation from meetings. We have proposed the Action Item Generation System (AIGS) which encompasses real-time and offline audio transcription , text preprocessing and segmentation, action item recognition. Harnessing the advanced text analysis capabilities of LLMs led to notable improvements in distilling pertinent action items for meetings.