Even though machine learning has become the major scene in dialogue research community, the real breakthrough has been blocked by the scale of data available. To address this fundamental obstacle, we introduce the Multi-Domain Wizard-of-Oz dataset (MultiWOZ), a fully-labeled collection of human-human written conversations spanning over multiple domains and topics. First results on multi-domain belief tracking and end-to-end modeling will be analyzed along with discussing open problems. In the second part of the talk, a new model for the multi-action dialogue management will be presented.
Paweł Budzianowski is a PhD student under the supervision of dr. Richard Turner and prof. Anna Korhonen in Dialogue Systems Group at Cambridge University. His research interests include multi-domain dialogue policy management, data collection for end-to-end dialogue systems and applied Bayesian deep learning. He received the best student paper award at ICASSP 2018 and the best resource paper award at EMNLP 2018. He gave a number of invited talks at academia and industry such as Apple, Google and Toshiba. He is also part-time researcher at PolyAI.