Conversational AI

Intelligent dialogue systems of the next generation for business

Voice assistants, chat and typing bots: they all enable interaction with machines and quick and easy access to information in the enterprise. However, existing conversational AI solutions largely disregard EU data protection standards and the specifics of European language diversity and cultural contexts. Many conversational systems also have limited reasoning capabilities or place information in context where none exists. As a result, their utterances are often less than natural or, in some cases, incorrect.

From chatbot to cognitive assistant

At the Fraunhofer IAIS site in Dresden, we develop intelligent dialogue systems that enable fast and efficient access to information via voice or text input. In doing so, we work with knowledge-driven language models.
The result: complete solutions for very different areas of application that meet data protection requirements and are tailored to the respective needs of the users - such as specific, German-language technical vocabulary or special requirements of the areas of application.

Conversional AI in interaction with knowledge graphs

We use modern AI technologies that combine Machine Learning and knowledge modelling. The special feature of this approach is fact-based AI: in the background of the dialogue systems are knowledge graphs, which combine generated knowledge from complex databases with specific domain knowledge from experts. This enables the dialogue systems to answer even complex queries correctly and in natural language.

Are you interested? We would be happy to arrange a non-binding demo or consultation appointment with you.

Benefits of our conversational AI solutions

Knowledge graphs

By combining it with knowledge graphs, our dialogue systems have access to all the data and information provided to it in order to learn based on it. Incorrect information or perplexity on the part of the dialogue systems are a thing of the past.

More info

State-of-the-art components

From speech recognition to question answering with knowledge graphs to speech synthesis – our offering covers the full range of Conversational AI.

European data protection standards

From Europe, for Europe: Our technology follows strict data protection standards and can therefore also be used in sensitive areas such as medicine without any problems.

Multilingualism

Previous language models and dialogue systems are specialised for the Chinese and American markets. Our Conversational AI is the best German language-based technology on the market and easily adaptable to other languages in the European region.

Simple scalability

Our solutions are quickly ready for productive use and can be easily adapted to your industry, your use case, and your technical vocabulary.

Application examples

Medicine

Medical devices, equipped with a voice assistant, are controlled by the medical staff via voice input and gestures. In combination with methods of knowledge extraction from medical documents, doctors receive important support in making diagnoses.

Business

Our Conversational AI solutions can support business processes or help to quickly find and access the right information in corporate or media databases.

Automotive industry

Integrated in the car, the dialogue system can answer the driver's questions and provide information about the vehicle or the surroundings.

Large AI language models as the basis for conversational AI

Large language models are a key technology in Artificial Intelligence. They recognise, produce, translate, and process language at a level that is difficult to distinguish from human-produced language. This enables them to perform increasingly complex tasks, such as conducting customer dialogues or writing software.

Learning contexts with neural networks

A language model maps how words relate to each other. "Car", for example, has a relationship to "traffic". The AI software learns such relationships based on knowledge inputs through texts from websites and books and with the help of neural networks. The more artificial synapses a language model has and the more text it has "read" during training, the better the programme can figure out how a sentence needs to be completed.

Data-based business solutions for European needs

The economic potential of new big language models is enormous. The most successful AI language models still come from the USA and China, such as the chatbot ChatGPT. These are often not available to the free market, or only to a limited extent, and are only used for the English and Chinese languages. Due to the rapidly growing importance of large AI language models, there is an urgent need to ensure technology and data independence as well as innovation and competitiveness in Europe. In the OpenGPT-X project (see below), Fraunhofer IAIS is working on a European answer to the American and Chinese large language models.

Companies benefit from the new technology

Interacting with a large language model in one's own native language, for example German, then comes very close to natural language and leads to enormous time and cost savings in companies. The principle is simple and the implementation time for companies is short: "The data and documents are uploaded into the AI system, and then you have a conversation with a virtual language assistant as if a human had read through it in no time at all," says Dr Nicolas Flores-Herr.

Telecommunication network above Europe viewed from space with connected system for European 5g LTE mobile web, global WiFi connection, Internet of Things (IoT) technology or blockchain fintech — © NicoElNino - stock.adobe.com

Digital sovereignty for Europe

About OpenGPT-X

The OpenGPT-X project partners build and train large-scale AI language models to drive innovative language application services for the European economy. Fraunhofer IAIS and IIS are leading the project.

Via the open Gaia-X infrastructure, companies will be able to use and share data and services free of charge, in multiple languages and according to the highest European data protection standards. The services offered will serve as a basis for developing products and processes with a wide variety of language functions (e.g. chatbots, digital assistants or personalised media reports). The cooperation between science, business and technology is funded by the Federal Ministry of Economics and Climate Protection.

To the press release (in German)

About the LEAM initiative

In January 2023, the German AI Association published a feasibility study on the topic of "Large AI Models for Germany" at the LEAM conference, on which researchers from Fraunhofer IAIS collaborated. It analyses the potential of large AI language models and outlines the establishment of an AI ecosystem for the development and use of large AI models in Germany and Europe. A key element for this, according to the study published by the KI Bundesverband, is the establishment of a dedicated infrastructure of high-performance computers designed for AI. LEAM is an initiative of the KI Bundesverband and leading representatives from industry and research, including Fraunhofer IAIS, which promotes the development of large-scale AI models.

To the news (in German)