Céline Brault, Community Manager KPN
Speech to text API Contexta

According to research by Google, 61% of people searching for products on their mobile devices want to be able to call the companies that show up in their search results. Overall, 52% of customer interaction takes place by phone call.
Even in a world where customers prefer self-service and millennials are turning to text-based communication, the telephone remains a vital customer service channel.

However, for data-driven companies, the potential of unstructured data such as voice calls can't always be fully explored for the following reasons: 

  • Voice calls take place using speech, rather than the text with which big data and machine learning algorithms already work well.
  • The customer service agent is focused on providing great service and, rightly, that shifts mental focus away from taking accurate notes.
  • Generally, humans have to listen to call recordings in real-time and so they can be a time sink.

Now, though, there’s an API in the KPN API Store that solves that problem!

Contexta uses machine learning to uncover insights in voice calls and turn them into a rich source of knowledge.
With the Contexta API, you can apply big data analysis to all of your customer interactions. Aside from telephone calls, you can also feed it with email, social media and any other text-based customer service history.
The result is that you’ll be able to rapidly categorise and label customer service calls and other interactions. From there, you can detect trends, track the topics that are most important to customers and your business, and analyse customer satisfaction trends over time.

Turning speech into text

Turning voice calls into useful data is a three-step process:

  1. Transcribe the call from speech to text.
  2. Analyse that text.
  3. Turn that analysis into business value.

Throughout the history of computing, humans have had to learn to think and communicate like computers. If you want to instruct a computer, you must learn a programming language. If you want to work with a computer, you must learn how to use the relevant applications.
With tools like Contexta, we’re finally able to make computers understand humans. The first two steps –– transcription and analysis –– would have been largely impossible ten years ago. 
Today, Contexta can take the audio from a customer service call and transcribe it at around twice the speed of the call itself. The resulting transcript is an accurate textual representation of the call that even identifies each participant.
The really hard part though is to get past all that ambiguity, inferred meaning and cultural context that we humans throw into conversations without realising. This is where Natural Language Understanding comes in. Contexta takes the transcript and analyses it to understand the content of the conversation. With that understanding, it can categorise the call, track how the customer’s sentiment changes throughout and look out for industry-specific keywords. And as a European company with its HQ in Amsterdam, Contexta works with German and Dutch as well as English.

How Contexta can help you

Let’s look at an example of what Contexta can do for you. Imagine you run a small contact center. Each day, hundreds of people call your contact center with sales questions and support queries. To help train and motivate your team of agents, you want to measure three things:

  • Time to resolution
  • Customer happiness
  • Creation of cross-sell opportunities.

You could do this manually. A team lead could spend a few hours each week listening to call recordings and checking, by hand, if the customer was left happy and that the agent did their best to cross-sell other products. With hundreds of calls per day, the team lead could only ever dip into a few calls per agent. And it can be boring work; not only could the team lead’s morale suffer but they might miss important details as their attention wanes.
With Contexta, you could check every single call and without taking up staff time. Contexta can look out for key phrases and words in calls, such as those that indicate a cross-sell or a successful resolution to the call. It can even generate transcripts so that human managers can quickly and easily review call content without having to listen to them in real-time.

Try the Contexta API for free via the KPN API Store today and build a deep understanding of customer service interactions directly into your existing customer management platform or into your own customer tools.

Watch also the video of Pranav, one of our developer, presenting the API!