Skip to content
Oloff Biermann edited this page Dec 31, 2022 · 17 revisions

Discribe - a Product of Workplace Futurists

Welcome to the Wiki Site of Discribe - a Product of team Workplace Futurists. This is a UBC CPSC319 course project created for HSBC.

Sections

Project Goal Achievements

List below is Created Based on the Original CS 319 HSBC Project Briefing (Ticks means acheived - We acheived all goals). (Text in brackets is comments and explanation of our actual implementation).

Component 1

  • A dialer/recorder agent that is invited to meetings via the regular outlook calendar meeting invitation.
  • Once the meeting time occurs, the scheduler will launch a program (Twillio) that dials into the audio-based call and starts the recording of the audio blob meeting.
  • Upon the meeting completion, the voice blob recording is stored in a common folder for transcribing. (We use a stream approach instead, but acheived the same feature).
  • This component can schedule meeting invites that need transcription within ½ hour of the originating invite.
    (our time is almost instant, much less than 1/2 hour).
  • When meetings are called, the system will pre-check that all participants are registered in the voice print system (component 3). Those users not registered will be notified via email with an registration URL.

Component 2

  • A Transcribe Agent backend job that may run in a batch mode whereby, it opens each recorded file and transcribes the contents into text. The files are saved in another directory for future distribution. (We actually use stream to process)
  • The service will leverage the Microsoft Cognitive Services library and transcribe the English language meeting and distinguishing the various participants.
  • The service will be tuned and adjusted to accommodate the various accents of participants.
  • After transcription, the text file is emailed to the chair person of the meeting.

Component 3

  • A voice print user registration that identifies each potential participant. This module will record a sample voice for the user that can be used by the transcribing agent to identify the user.
  • Each registered user will have an entry in the Voice recognition database.

Stretch Goals

  • Create a third module that will deliver the transcribed file back to the meeting chair (the owner of the Outlook calendar invite).
  • The transcribe agent can manage invites from multiple timezone and the scheduler can accommodate accordingly.
  • Repeating invites can be addressed by the system.
  • Maintenance module for the voice recognition component.