AICoE Project

  • Data Representation and Learning for Dialogue System

Project Name

Data Representation and Learning for Dialogue System

Project Goal

• Enabling conversational AI in heterogeneous environments, building virtual agents or assistants for smart

campus and hospital, enhancing deep learning theories, and facilitating explainable AI

• Enriching and integrating dialog agent, speech recognition, natural language understanding and generation


Project Description

1.Statistic Explainable Deep Model

*Spoken dialogue system

*Online speech recognition

2.Online Speech Recognition

*Compress previous speech

*Generate transcription immediately

3.Online Inference

*Modified one step constrained beam search

*Compressive memory processing

4.Behavior Cloning

*Auxiliary tasks for imitation learning

*Causal confusion reduction