When you don’t know what you don’t know, audio analytics can help

​In the popular TV show CSI, forensic investigators painstakingly solved crimes by pouring over crime scene evidence. Over the course of the hour-long show, using brilliant detective-work, they’d always find the missing pieces and crack the case.

Of course, the detectives of CSI may have made it look easy, but in real life, it isn’t always so. If an investigator’s not sure what he’s looking for, identifying the right places to look, then finding the few hidden nuggets of evidence that will lead to a conviction can be immensely difficult and time consuming. 

This can be especially true with 9-1-1 audio recordings. Recorded audio can be invaluable for investigations. But unlike physical evidence, it can’t be visually inspected, or picked up off the ground. The investigative clues lie hidden beneath the surface, in the spoken words themselves. These clues can be difficult to get to, especially when there’s no time/date reference, or if the incident is complex (occurred over some distance and time). So, in this circumstance, how does an investigator find the tiny needles in the big haystack, when he doesn’t know what he doesn’t know in the first place?
Audio analytics can provide an answer to this perplexing problem. Audio analytics has been in use for many years in commercial contact centers to uncover hidden business insights and to assess agent performance. Only recently, though, has it been developed to meet the specific needs of public safety. The biggest difference between audio analytics for enterprise environments and audio analytics that is purpose-built for 9-1-1 environments is that the latter delivers extremely accurate and fast results, even if you don’t know what you’re looking for. 

A good case-in-point is the Inform audio analytics solution which NICE will be demonstrating for the first time at NENA 2015. The solution is speech-to-text based (rather the phonetics-based) and it literally indexes every word found in 9-1-1 audio recordings – which means it can reliably detect every possible occurrence of a word or phrase. All word matches are indexed along with the estimated confidence level that the match is correct.  This methodology means that an investigator does not need to know specifically what words he’s looking for ahead of time.

NICE’s solution also indexes words in an ‘inverted-index.’ Rather than listing each call and the key words or phrases contained within the call, the inverted index uses the key word or phrase as the main reference point, which it maps to recorded calls that contain the key word or phrase. 

A detective can search 9-1-1 audio that spans an extended timeframe, by any word or phrase that could be related to the investigation. For example, if the investigator is working on a robbery case – say a store was broken into during the night – he might want to search for the name of the store, the location, etc. If a description of a suspect was available, the investigator could use that descriptive terminology to find any relevant calls. This investigative approach could even lead to finding more witnesses, or earlier calls involving other break-in’s that the suspect could have been involved in. 

While you probably won’t see audio analytics featured on a CSI episode re-run, it’s quite possible the technology will become a regular ‘extra’ in 9-1-1 centers. It’s a revolutionary approach to audio investigations whose time has come.
Share this:
Twitter LinkedIn Facebook Email