Goal-oriented dialogue state tracking by using BERT

Dialogue state tracking (DST) is a core component of virtual assistants such as Alexa or Siri. To accomplish various tasks, these assistants need to support an increasing number of services and APIs. In this work, we propose a GOaL-Oriented Multitask BERT-based dialogue state tracker (GOLOMB) inspired by architectures for reading comprehension question answering systems. The model «queries» dialogue history with descriptions of slots and services as well as possible values of slots. This allows us to transfer slot values in multidomain dialogues and have a capability to scale to unseen slot types. Our model achieves a joint goal accuracy of 53.97% on the SGD dataset outperforming the baseline model.