Automated MoM Generation with IoT-Based Audio Capture and Machine Learning Integration

Abstract

Businesses together with academic institutions benefit from
efficient documentation of their meeting proceedings in our
current fast-paced commercial and academic environment.
Traditional handwriting for note-taking usually fails to produce
accurate results along with causing prolonged delays. This
system addresses the challenges through automated Minutes of
Meeting (MoM) generation by processing voice recordings
through machine learning (ML) approaches. An IoT-based system
utilizes current STT algorithms with real-time voice capture
through IoT devices to generate audio transcription results. The
NLP system processes textual data to identify important meeting
points along with decisive decisions and task-related items.
A three-part approach served as our suggested plan. The meeting
audio gets recorded through IoT microphones which transform
the voice into text content using pre-trained speech recognition
software. The text transcription process requires initial cleansing
by eliminating verbalization fillers in combination with
diarization breakdowns before tokenizing the document content.
The third operational stage implements ML-based extractive and
abstractive summarization algorithms which extract crucial
decisions and actions from the meeting. The text summarization
and topic modeling process relies on Transformer-based models
including BERT and GPT. A specialized approach for action-item
detection runs as part of the process to highlight essential duties
and obligations.
The solution leads to accurate and fast MoM creation through minimized human involvement and boosted meeting efficiency.
The methodology demonstrates great value when big teams
along with organizations need to document their multiple regular
meetings. Research demonstrates that our system achieves high
accuracy when detecting action items along with summarizing
decisions thus shortening the period needed for human-made
MoM.