Today Nexidia announced a new solution that allows call centers to use speech analytics in real time.
Normally this would have been very cool by itself (despite what you see in movies the technology we have cannot yet select a phone conversation from among millions – or even thousands – based on two or three keywords very accurately), but when I was being briefed on the solution I picked up on two additional things that I think are worth mentioning: the patented technology, and their desire to use it in new areas.
Briefly, speech analytics works by converting sounds to words. This is, usually, done word-by-word – and you can already imagine why real-time speech analytics is almost impossible to do. Traditional speech recognition techniques that interpret words one-by-one work at 3-4 words per minute, which means they can analyze one hour of audio in about 15-20 minutes. Nexidia found a way to accelerate the using a patented phonetic interpretation model that, just like is used to teach kids to read faster, speeds recognition up by magnitudes. What used to be done in a 4:1 ratio can now be done in a 700:1 ratio; what used to take 15 minutes can now be done in seconds. Behind the speech recognition engine there is a very powerful, very open analytical engine that was created specifically to handle real-time analytics.
Their announcement for today was a very cool application that I got to preview last week that takes clues from the conversation (key words, phrases, contextual information, etc.) and uses that to pull data and knowledge from the corporate systems and the knowledge base to assist the agents automatically as the conversation goes on. It is dynamic, so it is continuously monitoring the conversations and providing more assistance. I saw a couple of demos and they looked really good very interesting. When deployed, it should help call centers cut down on Average Handle Time, increase Customer Satisfaction, and promote Employee Morale. The results from the first clients showed significant improvements already.
So, why does it matter?
The only thing standing between Social Business becoming a reality is the lack of good analytical tools that can operate in quasi-real-time to help businesses make better decisions. Actionable insights, as I wrote in my introduction to the Roadmap to SCRM, is the key outcome of implementing SCRM. It is the value that business derive from interacting with customers across so many partners at the same time. Valuable information on how to improve the business and better serve their customers — even in real-time – is the reason companies become Social Businesses.
During our conversation I deviated the talk from speech recognition to the embedded analytical engine and they confirmed that is very flexible and dynamic, that it can take input from anywhere – even social media channels (they are evaluating how to use it differently in the future).
My light bulb went off — what about putting the speed of processing of their analytical engine, the ability to support cross-channel, and the very cool app they have and turn it into a social business analytical solution?
We will have to wait, but for now the call center can deploy a very cool solution for real-time speech analytics.
____________ Disclaimer: I have not been paid for writing this, nor will I get paid in the future. This is my opinion, my position, my vision – my whatever-you-want-to-call it. I truly want to see this tool moved to Social Media without loosing the power of the speech analytics, and that is the purpose of this entry. Nexidia is not a client, nor do I have any friends or relatives that work there or for one of their contractors. Seriously, get a demo because this is cool stuff – yes, kinda like rocket science. As usual, do your due diligence and don’t take my word to be worth more than an opinion. We all know what opinions are…
ThinkJar - The Blog (Archive)! 
Esteban,
Thanks for sharing your excitement regarding speech analytics, and your ideas on using Nexidia’s analytical engine to facilitate social business. There is a lot to gain from analyzing cross-channel interactions and optimizing cross-channel customer experiences.
Nexidia offers great software. No doubt about that. I tested a previous version of the software, and I too was impressed. Nexidia is not the only software vendor offering real-time speech analytics and agent assistance though. There are other vendors that combine a phonetic approach for speed and fuzzy search/matching (i.e. the ability to recognize unknown words in the audio stream or misspelled words in the back end systems) with a word based approach.
A word based approach, as opposed to a purely phonetic approach, enables you to find trends and hot and breaking topics in your call recordings without having to define a particular search string first (i.e. without having a priori knowledge of what you are looking for). This can be a great asset for call centers to explain call spikes that cannot be explained easily with the knowledge at hand. A word based approach also enables you to do conceptual matching, i.e. mapping customer questions with documents, interactions etc. on the basis of a semantic understanding of both the customer request and the information in your back end systems.
Then how to decide which software/approach to use? If you know what you are looking for (i.e. you can define your search terms or a boolean search), go for a phonetic approach. It is quicker than a word based approach, and it enables you to do fuzzy search/matching. If you don’t always know what you are looking for (i.e. if you want the system to automatically cluster customer interactions without human supervision to explain call spikes and the like), or if you want to perform a deeper level of conceptual matching (to find synonyms or conceptually related documents), a word based approach or a combination of both approaches may offer a better solution.
As an aside, word recognition is faster than you assume. Nuance Dragon Naturally Speaking for example offers dictation software that delivers excellent word recognition in near real time, up to three times faster than most of us type.
I share your enthusiasm, Esteban. Nexidia offers great software. I am excited about their new release, and the potential of speech analytics for analyzing and improving customer interactions. I can’t wait to get my hands on the new version of the software.
Kind regards,
Christophe.
LikeLike