Task-Oriented Dialogue as Dataflow Synthesis
This addresses the challenge of handling complex dialogues for AI assistants, though it is incremental in applying dataflow concepts to dialogue state.
The paper tackles the problem of representing complex user intents in task-oriented dialogue by introducing a dataflow graph approach, which improves representability and predictability, as shown by matching state-of-the-art models on MultiWOZ and enhancing performance on the new SMCalFlow dataset.
We describe an approach to task-oriented dialogue in which dialogue state is represented as a dataflow graph. A dialogue agent maps each user utterance to a program that extends this graph. Programs include metacomputation operators for reference and revision that reuse dataflow fragments from previous turns. Our graph-based state enables the expression and manipulation of complex user intents, and explicit metacomputation makes these intents easier for learned models to predict. We introduce a new dataset, SMCalFlow, featuring complex dialogues about events, weather, places, and people. Experiments show that dataflow graphs and metacomputation substantially improve representability and predictability in these natural dialogues. Additional experiments on the MultiWOZ dataset show that our dataflow representation enables an otherwise off-the-shelf sequence-to-sequence model to match the best existing task-specific state tracking model. The SMCalFlow dataset and code for replicating experiments are available at https://www.microsoft.com/en-us/research/project/dataflow-based-dialogue-semantic-machines.