Emotion-Aware Conversation System Using a Conditional Transformer and Reinforcement Learning

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Emotion-Aware Conversation System Using a Conditional Transformer
        and Reinforcement Learning"

by

ZHANG Chi

Abstract:

Conversational systems, also known as chatbots, interpret utterances from 
users and responds with human-like natural language have long been an 
important research topic in the field of natural language processing. 
However, many generic conversational systems fail to factor in the 
critical role that emotion plays in human communication, resulting in 
mediocre responses. For better human-computer interaction, conversational 
systems should be able to understand human emotions and respond 
accordingly. In this study, we propose a novel solution to build an 
emotion-aware conversational system that has high performance in fluency, 
relevance, and empathy in a daily conversation context. We used 
Transformer as the basic skeleton of our system and added an emotion 
encoder to understand the user's emotion and control response generation 
with appropriate emotion. According to our experiments, our model can 
provide a more emotionally positive response compared to baselines.


Date            : 8 May 2021 (Saturday)

Time            : 15:10-15:50

Zoom Link:
https://hkust.zoom.us/j/92133587886?pwd=N3orUFNBNFpSVjZiV1JTM0Y5TVNZZz09

Meeting ID      : 921 3358 7886
Passcode        : 663031

Advisor         : Prof. ZHANG Nevin Lianwen

2nd Reader      : Dr. SONG Yangqiu