JOSEPH — Language Technologies With Focus on Data Ownership Facilitate Smart Living

Why do we need this new technology?

Speech technologies are an important basis for human-machine communication, enabling simplified interaction with services and sources of knowledge along with automated control of equipment and machines. Voice assistants like Alexa and Siri make daily life easier for millions of people.

Ever since ChatGPT was launched, the capabilities of language models that automatically generate text and images have garnered increasing attention. Companies based in the U.S. are especially strong in developing these technologies. However, they face criticism for collecting and processing sensitive user data. For data protection and privacy reasons, many companies and organizations shy away from using “external” artificial intelligence (AI).

Amid the new situation, in which large language models constitute a disruptive new base technology, the JOSEPH project aims to address industry demand for innovative speech solutions made in Europe to ensure the security and ownership of users’ data while at the same time driving technological advances in this field. Researchers at the Fraunhofer Institute for Integrated Circuits IIS and the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS are working to develop these kinds of AI-driven speech technologies that also comply with data protection and privacy requirements.

Who will benefit from the new technology?

There are a broad range of potential applications for voice technologies and artificial intelligence. In industrial production, they considerably simplify human-technology interaction and accelerate workflows. In automotive applications, they improve interaction between vehicles and drivers and make autonomous driving possible. Financial and administrative processes can be streamlined. Healthcare professionals can operate equipment without touching it and retrieve information. People with disabilities can receive assistance with day-to-day tasks for greater independence.

How does the new solution work?

To be able to offer competitive solutions, the Fraunhofer institutes are focusing their activities on developments in the areas of automatic speech-to-text (STT) and text-to-speech (TTS) technologies, alongside natural language comprehension, language generation and dialogue management. In particular, they have realigned their work in the areas of language comprehension and dialogue management on the basis of language models.

What makes the project unique?

JOSEPH strives for a new degree of quality in human-machine communication, far beyond the semantic capabilities of current systems. The technology guarantees digital sovereignty while also meeting German and European standards of data security.

Why is the Fraunhofer Future Foundation supporting this project?

The Fraunhofer Future Foundation is funding the JOSEPH project to serve the great demand for AI-based voice technologies developed in Europe. The project will be focusing at first on applications at companies in Germany and elsewhere in Europe, as there are concrete collaborations in these areas. Then, as the second step, the findings and developments can be translated to applications for end users. These technologies are scalable and constitute a valuable alternative to the currently widespread language solutions. The goal is to ensure that this innovative AI-based technology achieves international visibility and availability.