Ever wonder how you do audit in a large call center?

Here's the basics:

First, most modern call centers are using Voip. It makes it easy to setup, and even better
makes it easy to monitor. Also call-centers tend to be in places like india, or phillipines
serving the US or Europe markets. So VoIp plays a role in getting the traffic from one
country to another.

There are several pieces to this puzzle:

First, you need to capture the raw data:
So to do that, we need a "tee" off the ethernet channel that is connecting to the VoIP gateway.
This will allow us to see all the VoIP IP Traffic that is going back and forth.

It's sometimes called 'port mirroring', 'port monitoring', 'Roving Analysis' (3Com), or 'Switched Port Analyzer' or 'SPAN' (Cisco).

http://wiki.wireshark.org/CaptureSetup/Ethernet
http://www.cisco.com/en/US/products/hw/switches/ps708/products_tech_note...

Now that we have traffic, a good way to "capture" the voice is Oreka
http://oreka.sourceforge.net/

Oreka lets you capture the packets, and convert them to audio files. It actually consists
of three parts.

- OrkAudio: This is the workhorse that processes the calls and does the actual recording
- OrkWeb: This is the XML based Web U/I to access and manage the system
- OrkTrack: This is the master database (MySQL) that records the call records, metadata, etc.

Oreka lets you get a basic system up quickly.

The next issue, is setting up a system of audit, and to connect and understand the calls into transactions,
customers, call-agents, and supervisors.

For a good GUI on top, I'd switch to a Rails App. This lets me do several things:

1. I can tie agents to calls
2. I can allow agents access to review there own calls
3. I can allow supervisors to view/listen to calls for agents reporting to them
4. I can let audit department audit calls as appriate.
5. Ease of interfacing existing call center apps togeather

Now that we have a phase 1, and phase 2 completed, lets add a phase 3.
One of the emerging technologies is Speach to text.
Using the latest SDK's from Dragon Speaking, we can get very high 99% plus on our agents,
and not bad recognition on our clients/customers calling in. This will allow audit to search
for key words, such as scam, theft, etc that would audit to be better aware of things going
on.

The idea way of implementing the whole system is break things down to separate VM's in a VmWare system.

1. OakCapture - 1 VM per major Trunk (One Per - Multiple DS1, One DS3)
2. OakGui - 1 VM For system - For reference/admin control only
3. Database - In a call center, a Oracle or DB2 may be better suited, depending on the company
4. VoiceCenterManager - A Rails Server using Nginx/Passenger that provides interfaces to agents, supervisors, audit,
and managers. Number of VM determine by staff size
5. Recognition Engines - A separate VM that is running a Ruby Agent tied to a Dragon Speaking SDK allowing speaker dependent and Speaker Independent Speech to text conversion. This allow for the creation of easily searchable as well
as readable results of the calls in text form. Using a rack of blades, and VmWare we can realtime convert all calls to text
index them, and make them searchable.

Note that each call center is different but leveraging Rails, and Ruby a AGILE and fast solution can be combined that
gives the best in Audit and accountability.