Structured Attention for Unsupervised Dialogue Structure Induction
This addresses the problem of automatically inducing dialogue structures for computational linguistics, dialogue system design, and discourse analysis, representing an incremental improvement over existing methods.
The paper tackled unsupervised dialogue structure induction by incorporating structured attention into a Variational Recurrent Neural Network (VRNN) with discrete latent states. The result showed that on two-party dialogues, the model learned semantic structures similar to generation templates, and on multi-party dialogues, it learned interactive structures distinguishing speakers without human annotation.
Inducing a meaningful structural representation from one or a set of dialogues is a crucial but challenging task in computational linguistics. Advancement made in this area is critical for dialogue system design and discourse analysis. It can also be extended to solve grammatical inference. In this work, we propose to incorporate structured attention layers into a Variational Recurrent Neural Network (VRNN) model with discrete latent states to learn dialogue structure in an unsupervised fashion. Compared to a vanilla VRNN, structured attention enables a model to focus on different parts of the source sentence embeddings while enforcing a structural inductive bias. Experiments show that on two-party dialogue datasets, VRNN with structured attention learns semantic structures that are similar to templates used to generate this dialogue corpus. While on multi-party dialogue datasets, our model learns an interactive structure demonstrating its capability of distinguishing speakers or addresses, automatically disentangling dialogues without explicit human annotation.