Towards Emotion Intelligence in Neural Dialogue Systems

• Author / Creator

  Huang,Chenyang

• Dialogue systems, also known as Conversational Agent (CA), are designed to mimic coherent conversations with humans. Most conversational agents are specialized for a specific domain such as travel booking and are typically finite state-based or template-based. Open domain dialogue systems have seen a growing interest in recent years thanks to neural dialogue generation systems, based on deep learning models.

  These systems basically learn to predict the words and the sentence to
  respond based on the previous utterances. However, while such a system can generate grammatically correct and human-like answers, the responses are often generic and non-committal instead of being specific and emotionally intelligent. In this work, the objective is to tackle two main problems that are essential towards building emotionally intelligent chatbots: “How to detect the emotions expressed by the human accurately?” and “How can a chatbot express an emotion?”

  We propose a Neural Network model which is dedicated for emotion recognition. It combines multiple recent advances in semantic and emotional feature representations. Our experiments show that the proposed model outperforms the current state-of-the-art models by a large margin. We then develop a hierarchical variant of the model and get outstanding result on the SemEval2019-Task3 shared task.

  In order to design a more reliable/generalized emotion detection system, we also collect a dataset from scratch which is more than 10 times larger than the current largest emotion dataset that is publicly available.

  In order to generate specific emotions in open-domain dialogue environment, we propose a total of seven models that are all based on a widely used neural dialogue generation framework. The results indicate that all the models are able to tackle the task equally well, and we compare the models in terms of accuracy and computation costs.

  We also reflect on the problems of anticipating expressed emotions but an interlocutor and the problem of determining the appropriate emotion to express in a generated response.

• Subjects / Keywords

  • Emotion Intelligence
  • Machine Learning
  • Dialogue System
  • Neural Networks

• Graduation date

  Fall 2019

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-p453-d634

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.