S4CS – Speech Assistance for Citizen Services

Interaction with authorities is often complex and lengthy. Can long waiting times and the time-consuming filling out of applications be avoided? Fraunhofer IDMT in Oldenburg and Fraunhofer FOKUS in Berlin are addressing this question in the »Language Assistant for Citizen Services« project.  

The combination of intelligent speech recognition from Fraunhofer IDMT with an engine for semantic processing from Fraunhofer FOKUS is intended to simplify digital exchange with public authorities. The goal of the »S4CS« project is to develop a demonstrator that supports citizens in requesting official services. The first scenario chosen was to apply for parental benefits via language assistant.

Enabling digital, simple and intelligent communication in natural language within such a process brings with it a multitude of added values for citizens and authorities. In addition to saving time, reducing complexity and increasing accessibility, the project is also highly relevant in practical terms as it contributes to the goals of the Federal Government’s digitalisation programme (Onlinezugangsgesetz/OZG). These demand, among other things, that all municipalities must offer their administrative services online by 2022.

Realization

Based on deep neural networks, Fraunhofer IDMT has developed a reliable speech-to-text analysis that is insensitive to background noise and scalable. Local data processing ensures the confidential handling of voice data in compliance with the highest data protection standards. The Constraint Handling Rules-Engine (GoCHR) developed by Fraunhofer FOKUS is used for understanding and answering citizen queries. The easy-to-understand constraint rules can be adapted quickly and flexibly for new administrative services. The use of machine-learning methods is also aimed at for language generation (text-to-speech).