Transcription in noisy call centers
- On July 18, 2018
- ASR, contact center, transcription
Today, modern call centers are deploying sophisticated software components to help them improve customer experience, increase productivity and comply with new regulations. One of the enablers for these new capabilities is the requirement to have a reliable and accurate analysis and transcription of the agent’s voice. Such analysis and transcription is used for many purposes like:
- Quality and performance management.
- Verify the agents provides all the required information and complies with regulations like MiFID II and GDP.
- Identify the courtesy and mood of the agents.
A well known fact is that most call centers are far from being quiet locations, actually they are usually very noisy. The noise in the call center is generated by many sources like:
- Multiple agents that are talking simultaneously on the phone.
- Supervisors or team leaders that are shouting instructions.
- Internal communication systems (e.g. PA).
It is clear that these noises reduce customer satisfaction and also cause data leakage by exposing the customers to private and confidential data like commercial offers that are given to other customers. But, apart from this observation, do these noises have an impact on the analytics and transcription that is done on the voice of the agent?
The answer is positive. Even the most advanced analytics and speech-to-text engines might have some capability to disregard non-human noises but they cannot disregard voices of other people. They do not have the technology to distinguish between the voice of the primary agent and the voices of the supervisor and the nearby agents. As a result, the output of the analytic engines can be misleading. Let’s take a common example of an agent that is talking calmly on the phone and in the background we hear the supervisor shouting. In this example, the transcription will not be accurate and also the mood in the call will not be identified correctly.
How can this problem be solved? The best way is to prevent these noises in the first place, but since this is usually not a practical alternative, you can install an advanced noise-firewall software, like PBXMate for Call Centers, that will remove these background voices from the calls. PBXMate will simultaneously listen to all the calls, correlate their audio in real-time, identify the source of each voice and remove it from calls in which it causing noise. In addition to improving the quality of the analytic, by using PBXMate you will also improve customer experience and prevent data leakage from your call center.