Automated dialogues, be it via Siri, Alexa or Google Home, use chatbots and bot assistants that are based on systems for “Conversational Artificial Intelligence (Conversational AI)”. These chatbots and language assistants are trained using machine learning methods from texts and dialogues in the respective domain and language. For example, they learn from real dialogues how to order food in Italian. But the chatbot “trained” for one domain cannot simply be transferred to another domain or work well-suited in other languages. This would highly require training with dialogues for the specific domains and languages.

The project is aimed to examine methods for transferring conversational AI models across domains and languages, even with a limited number of dialogues or, in extreme cases, no dialogues at all of the target domain or target language.

We aim to advance the state-of-the-art in conversational AI by implementing a framework for fast and effective bootstrapping of dialogue systems for new languages and domains. We will exploit and build upon cutting-edge developments in natural language understanding (NLU), primarily in representation learning and transfer learning across languages and domains.

In the course of the project, the researchers will create AI agents for various domains and languages and make them availabke as "plug-in" modules for researchers and companies for future needs.