VoIP Colorizing Logger

This article describes a system for making useful transcripts of Voice over IP (VoIP) chats.

If you already have a VoIP based conferencing system that’s taking and mixing several users’ voice inputs and rebroadcasting the result, it would be handy to have an automated transcriber that could record the conversation in a useful format for later reference. Transcription results could be available on a website after the call, on a website in realtime, streamed over IM in realtime, or emailed out to participants as the call terminates.

A speech parser process is attached to each incoming line, and as each word is completed is appended to a shared buffer with a tag corresponding to the voice line being parsed. Results are colored per the voice line, and preliminary output could be as follows:

Hi. Jim here. Anyone else on this line yet?

Ted here. John is here too. Let’s begin.

Hi Jim.

Great. Are we a go for the presentation tomorrow?

We’ll need to update the Northeast numbers. They’re off.

Mary said she’d fix that tonight. I think we..

We’ll actually just be presenting West Coast numbers.

Oh.

Okay, great.

…

After the conversation has been completed, a third party, like a secretary, could go over the results and fill in which color corresponded to which name. This could be especially useful if subsequent conversations were to be stored in the same database – searches could then be performed, such as finding out what Ted said about Q3 numbers in the last few conversations. Having pre-filled “likely participants” available as a drop-down list selection could make quote assignment particularly easy. It is understood that this system might not work ideally for situations where multiple parties are on a single line. But as offices become increasingly virtualized, it’s more likely than ever that none of the participants are actually in a room together during a meeting.