Technology Preview - Broadcastify Calls Transcriptions

[blantonl]

Merry Christmas!

A technology preview is out. Immediate transcription of Broadcastify Calls. We're doing this for all fire dispatch channels in the Dallas Fort Worth Metroplex.

DFW Metro - All Fire Dispatch

All dispatch channels for most agencies in Dallas, Tarrant, Collin and Denton Counties in one single playlist

www.broadcastify.com

We're currently building out a complex pipeline and infrastructure for this, to include alerting and notifications, searchable database of transcriptions, and the infrastructure pipelines that will support this effort eventually across every Fire Dispatch, Fire Tac, Law Dispatch, and other non-sensitive channels and talkgroups.

The scale of this project is extensive and expensive. It's going to be freaking awesome.

 

[Deziel0495]

That's really cool! Also, is that a database RID I see there?

 

[Enforcer52]

I have been watching the feed and find the translation to be so much better than the other unnamed system, and controlled by RR.

Now this I can live with, and hope this turns out to be what we would expect in the scanning and streaming world.

Hopefully it can be applied to the public playlists that we have now.

 

[xicarusx]

Merry Christmas!

A technology preview is out. Immediate transcription of Broadcastify Calls. We're doing this for all fire dispatch channels in the Dallas Fort Worth Metroplex.

DFW Metro - All Fire Dispatch

All dispatch channels for most agencies in Dallas, Tarrant, Collin and Denton Counties in one single playlist

www.broadcastify.com

We're currently building out a complex pipeline and infrastructure for this, to include alerting and notifications, searchable database of transcriptions, and the infrastructure pipelines that will support this effort eventually across every Fire Dispatch, Fire Tac, Law Dispatch, and other non-sensitive channels and talkgroups.

The scale of this project is extensive and expensive. It's going to be freaking awesome.

That is neat, as a developer who has already built a similar infrastructure for 250+ audio feeds over the past 3 years. It is big bite to take, cost wise and complexity. Scaling from one system to many took some time and trial and error. As of today the system I have created is comfortably transcribing 2.1 million recordings a day. With only a handful of GPUs. Roughly 1500 a minute as of this moment and generating about 40 alert leads a minute on those transcripts.

Good luck with this endeavor.

 

[ffexpCP]

I wonder if this could expand to decode tones, then set feed or email alerts?