Let’s start with Rezan, one of our Speech Analytics experts.
Rezan has been working as Project Manager for Telecommunication Solutions at CreaLog for 8 years. He studied Computer Science at the Johann-Wolfgang-Goethe University in Frankfurt and Media Engineering at the University of Applied Sciences in Wiesbaden. At CreaLog, he designs Voice- and Graphical-User Interfaces and optimizes grammars for Nuance speech recognition in multiple languages. During his tenure at the Fraunhofer-Institut für Integrierte Schaltungen IIS in Erlangen, Rezan designed and implemented a Universal Transcoder Framework, a key component in the EU-Project EDCINE.
Together with my colleagues Dr. Christian Heinrich and Ralph Kynast, I'm currently implementing Speech Analytics at one of the largest insurance companies in Germany. There, current customer issues being handled at various hotlines are recorded and evaluated in near real time. My responsibilities within the project include optimizing our tools and workflows. The specification of the user interfaces, which we adapt exactly to the needs of our customers, is also part of my area of responsibility.
Speech Analytics provides deeper and otherwise hidden insights into internal business processes and customer dialogues. Based on this wealth of knowledge, our customers can take suitable measures to increase efficiency, improve service quality and customer satisfaction. Problems are recognized early thanks to Speech Analytics, additional business areas can be identified, their value developed and captured. This gives our customers an enormous competitive advantage. We of course know how crucial it is that all data collected and measures taken within the scope of Speech Analytics are always compliant with current data protection laws. Our system always ensures that sensitive data does not and cannot leave the company and that employee rights remain safeguarded. CreaLog relies on local, in-house and fully automatic recording and processing of the calls. In addition, further identification features, such as customer numbers, can be made anonymous.
The latest technological advances in the field of AI have triggered a real hype around this topic. This euphoria is partly justified. Deep learning offers promising possibilities for the automated assignment of conversations to topics of conversation or the recognition of emotions. However, such systems must be trained accordingly before being used by the customer. The process of manual evaluation and provision of sufficient test data by experts in the respective field are particularly cost-intensive and demanding. When certain conversation topics not contained in the test data arise, these cannot be recognized by AI. The same applies to completely new subject areas. For this reason, depending on the requirements, classic algorithmic methods also need to be considered.
As a principle, we orient ourselves towards our customers’ ideas and goals, so we can generate and offer the most suitable solution.