Big Data Analytics of voice by large call centers

Voice over IP (VoIP) and analytics tools are becoming each others need. Enterprises and large call centers are turning huge amount of telecom data into valuable intelligence to improve their business processes, generate new sales leads, increase customer satisfaction and gain insights into potential threats from their competitors.

VoIP is quickly becoming the first choice medium for telecommunications. That is why it should come as no surprise that VoIP brings the concept of big data analytics naturally. VoIP technology provides a good very fundamental method for capturing voice for recording from any call center setup which then can be piped into any other solution for more detailed processing on those calls. Here we'll discuss what different types of information we can get from processing of recorded calls.

Speech Analytics
Speech analytics (or Voice analytics) is the process of analyzing recorded calls to gather further information from the call. A call center may get hundreds (or even thousands) of calls per day. Each call vary from few seconds to few minutes. Some research shows that average call duration on all types of call centers is nearly 6 minutes per call. If you multiply it by the number of calls per day that becomes a huge amount of recorded audio data. So automated systems such as speech analytics can only give further insights to detect callers emotions, hold time, silence time etc.

In speech analytics systems, calls are also translated from voice to text and then complete transcript of the call is used for searching to find relevant topics. For example, a company can look to find calls with words of their competitor or product and then find the context in which they are used.

Emotion Detection
Speech analytics many also include analysis of topic of the call, the emotional words of the call, non-speech part of the call (e.g. slience, hold time etc). Since emotion is not an easy to detect part of the speech that is why it is sometimes handled separately. Emotion detection can be performed by specific strong language keywords and by analyzing pitch and tone of the caller. Sudden spike in tone and volume typically indicate an unhappy customer.

It's not possible for all types of call centers to use these expensive methods of analyzing call speech and emotions. But as hosted call centers are getting popular, these techniques are not far away from small call centers using the services of hosted call center solutions.