Automated Generation of Meeting Minutes using Deep Learning Techniques

Abstract

Meeting minutes are important to keep track of key decisions and agreements that were made during a meeting. It is crucial to document the topics discussed and the decisions made so they can be reviewed at the beginning of the next meeting as well as for future reference. Many companies while conducting meetings, keep paid employees to note down meeting minutes taking up valued time and resources. We offer a solution to make better use of available tools and technological advancements to help employees conduct effective discussions to boost a company’s productivity. Automated Minute Book Creation (AMBOC) uses machine learning to derive key information from important discussions. AMBOC is an automated system to create transcripts and minutes of a meeting with the added advantage of speaker recognition. The model we propose will be capable of transforming an audio file into plain-text using Deep Neural Networks (DNN), recognizing the speaker using Mel Frequency Cepstral Co-efficient (MFCC) as well as summarizing the meeting transcript into condensed minutes with the help of Transformers.