This meeting tool will be used to collect and parse audio from meetings for future analysis.
There are two different variations that can be implemented:
- A single stream of audio of meeting
- Multiple streams of audio of meeting
Components:
-
Audio collector:
- One single audio stream with buttons assigned to each person
- Multiple audio streams from different people that will be parsed together
-
Tagger:
- Each button will have a tagger associated with metadata for each person
- Each audio stream will have a tagger associated with the metadata for each person
-
Timestamps:
- Timestamps will be implemented when a button is pressed
- Timestamps will be recorded for pauses in a person's speech pattern
-
Transcription:
- The audio will be transcribed using the most successful Deepspeech Model
- The audio will be transcribed using the most successful Deepspeech Model
-
Post Processing:
- The audio will be broken down into intervals based on who was speeaking, using the timestamps from the tags
- The audio will be broken down into intervals based on pauses