2017 Poster Sessions : A Copy-Augmented Sequence-to-Sequence Architecture Gives Good Performance on Task-Oriented Dialogue

Student Name : Mihail Eric
Advisor : Chris Manning
Research Areas: Artificial Intelligence
Abstract:
Task-oriented dialogue focuses on conversational agents that participate in user-initiated dialogues on domain-specific topics. In contrast to chatbots, which simply seek to sustain open-ended meaningful discourse, existing task-oriented agents usually explicitly model user intent and belief states. This paper examines bypassing such an explicit representation by depending on a latent neural embedding of state and learning selective attention to dialogue history together with copying to incorporate relevant prior context. We complement recent work by showing the effectiveness of simple sequence-to-sequence neural architectures with a copy mechanism. Our model outperforms more complex memory-augmented models by 7% in per-response generation and is on par with the current state-of-the-art on DSTC2.

Bio:
Mihail Eric is a Master's student in Computer Science, concentrating in artificial intelligence. He has the great fortune of working in the Stanford Natural Language Processing Group, advised by Christopher Manning. Mihail's research interests span machine learning and natural language understanding, and he currently focuses on building goal-oriented dialogue agents. When he's not lurking in the hallways of Gates, Mihail enjoys reading, long-distance running, and salsa dancing.