Probably not unless it is done as a speech to text conversion. 3 Billion calls per day is a LOT of data. A years worth could be millions of Terabytes or more.
“The storage capacity of the Utah Data Center will be measured in “zettabytes”. What exactly is a zettabyte? There are a thousand gigabytes in a terabyte; a thousand terabytes in a petabyte; a thousand petabytes in an exabyte; and a thousand exabytes in a zettabyte.”
doing a quick google, i see some mention an average of 459 minutes per month for cell users. assuming double that to cover all home and work phone calls. rounding it to about 1,000 minutes/month... or about 33 minutes per day (rounding to 2000 seconds/day)
assuming 250m people in the US may use a phone, that would put the total audio volume around 500 billion seconds per day
allowing for a quality recording around 20 kpbs... the storage requirement would be about 10.24 quadrillion bits/day or 1,192,092 gb/day
allowing for 2 recording stations per state, 100 stations total, the per station recording would be about 12,000 gb/day. averaging across a 16 hr day, this would put the hourly recording requirement per station around 745 gb/hour
assuming at least 32 active drives to record incoming data for each station, this would put the load per drive to about 23.3 gb/hour ... or about 398 mb/min.. or 6.8 mb/sec.
drives today can write in excess of 100 mb/sec
drives in such a system would need to be changed about once a day. once swapped out they would be placed in storage for future reference, if needed
applying real-time voice recognition to these audio streams would produce a text file. the text file would then be parsed and indexed for phrases then rated across a multitude of categories. if it rated high enough, it would automatically be routed to the attending agents.
the name of the text file would be recorded in a database along with the time, date, duration, and call_file_id. this id would be used to map the call participants identity record to the call. the person_id table would also map to another table to record location information.
with a small bit of work... you now have a system that knows what was said .. between whom (numerous people in a call)... and where each person was while the call was taking place.
with a little bit of algorithmic magic, i could easily find a second order of associates using such a database... identifying the larger group.
and yes, i could easily put this system together ... given the funding.
therefore, i have no doubt such a system exists