Conversational AI

Intuitive interaction in B2B applications

Speech assistants are used in more and more areas of life, enabling intuitive interaction with technology, providing service and information. They are not only useful in everyday life, but also offer companies great potential for facilitating human-machine interaction and offering completely new services to customers.

We develop speech-driven dialog systems with a special focus on domain-specific knowledge for application in various fields of business and industry. Combining state-of-the-art components for Speech Recognition, Conversational AI via Knowledge Graphs and Speech Synthesis, our technologies address the concrete challenges and needs of enterprises and B2B applications.

Technologies »made in Germany«

Moreover, these technologies »made in Germany« ensure technological sovereignty, data can be stored and processed within secure data spaces. »Informed Machine Learning« methods, developed within the Fraunhofer Cluster of Excellence Cognitive Internet Technologies CCIT (Center for Machine Learning), make sure that the systems can even be trained on small data sets.

Our offer

  • We offer leading acoustic front end and speech recognition technologies for German and English.
  • Our dialog system needs less training data and can be used online and on-premise.
  • We offer a set of domain-adaptable state-of-the-art components for dialog systems, ranging from speech recognition to dialog management and speech synthesis.
  • We ensure digital sovereignty, data and technologies can be stored and processed in secure data spaces.
  • We offer services independent from the cloud to allow privacy e.g. when processing customer or business data.
  • We customize the systems for your specific requirements and help you leverage semantically structured knowledge from different data sources.
  • We offer certificated and standardized platforms, modules and datasets.

State-of-the-art components for B2B conversational AI

© Fraunhofer IAIS

1. Speech Recognition

Speech Recognition listens to spoken queries and converts them into written text. It can be adapted to domain specific terms and vocabulary and trained over domain specific knowledge for optimal performance.

2. Conversational AI

Conversational AI combines your domain-specific knowledge with machine learning approaches to enable faster access to your information. The concept of “Informed Machine Learning” leverages your data as background knowledge for a dialog system. Being stored in a machine-readable knowledge graph the system can easily access the data and search for suitable answers. In the end, the system uses a verbalization tool to ensure a natural formulation of the answers and to communicate with your users.

3. Speech Synthesis

Speech synthesis enables the dialog system to talk. It converts textual replies to spoken language and produces spoken output.

Use cases for conversational AI

© Fraunhofer IAIS
© Fraunhofer IAIS

Our Conversational AI platform enables access to your business data and databases via Natural Language Queries (e.g. spoken questions or chatbots). Our technology can be adapted to any domain-specific knowledge and is prototypically or productively implemented in various contexts:

  • The technology can support business processes and help to quickly find information in corporate or media archives.
  • Integrated into the car, the dialog system can serve as an interactive city guide, answering questions about points of interests.
  • Medical devices, equipped with a speech assistant, can be controlled by voice input. In combination with methods of knowledge extraction from medical documents, experts can receive crucial support in making diagnoses.
  • Dialog technology can save valuable time, for example when emergency teams use headsets to get fast and precise answers to life-saving questions.

Related Platforms and Projects:

  • SPEAKER, a speech assistant platform »made in Germany« (Start: April 2020)
  • ScaDS.AI, a big data competence center which allows collaborative exchange between domain and AI experts (Start: January 2020)