Sebastian Steindl

Brief description of the doctoral project

Chatbots are an important form of human-computer interaction and can be used in various customer service scenarios, for example. They can also combine several external service providers and thus enable complete holiday bookings, for example. However, if an organisation wants to train a chatbot for its specific use case, this often fails due to a lack of data, which is time-consuming and expensive to collect. My dissertation project is therefore concerned with synthesising this training data. On the one hand, I am looking at how this data can be generated artificially and, on the other hand, how helpful it is for training chatbots. I use both publicly available data sets and real company data. Both rule-based and data-driven approaches are used as methods. Large language models in particular open up many possibilities.