A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialogues
This work addresses dialogue generation for conversational AI, but it is incremental as it builds on existing neural network architectures with latent variables.
The authors tackled the problem of generating dialogue responses by proposing a hierarchical latent variable encoder-decoder model to capture complex dependencies in sequential data, and experiments showed it improves upon recent models in automatic and human evaluations.
Sequential data often possesses a hierarchical structure with complex dependencies between subsequences, such as found between the utterances in a dialogue. In an effort to model this kind of generative process, we propose a neural network-based generative architecture, with latent stochastic variables that span a variable number of time steps. We apply the proposed model to the task of dialogue response generation and compare it with recent neural network architectures. We evaluate the model performance through automatic evaluation metrics and by carrying out a human evaluation. The experiments demonstrate that our model improves upon recently proposed models and that the latent variables facilitate the generation of long outputs and maintain the context.