Bing Liu's Lifelong and Continual Learning Research Page

SIGIR-2022 Tutorial Slides: Continual Learning Dialogue Systems - Learning during Conversation

Two-Sentence Summary

The tutorial focuses on the new research topic of building the next-generation dialogue systems (or chatbots) that can continuously and interactively learn from end-users during conversation (post-deployment) to become more and more powerful over time. We will give the motivation and background of the topic and discuss existing techniques and open challenges, which we believe, will shape the future of dialogue systems development.

Abstract

Dialogue systems, commonly known as Chatbots, have gained escalating popularity in recent years due to their wide-spread ap- plications in carrying out chit-chat conversations with users and accomplishing various tasks as personal assistants. However, they still have some major weaknesses. One key weakness is that they are typically trained from pre-collected and manually-labeled data and/or written with handcrafted rules. Their knowledge bases (KBs) are also fixed and pre-compiled by human experts. Due to the huge amount of manual effort involved, they are difficult to scale and also tend to produce many errors ought to their limited ability to understand natural language and the limited knowledge in their KBs. Thus, when these systems are deployed, the level of user sat- isfactory is often low.

In this tutorial, we introduce and discuss methods to give chat- bots the ability to continuously and interactively learn new knowl- edge during conversation, i.e. “on-the-job" by themselves so that as the systems chat more and more with users, they become more and more knowledgeable and improve their performance over time. The first half of the tutorial focuses on introducing the paradigm of life- long and continual learning and discuss various related problems and challenges in conversational AI applications. In the second half, we present recent advancements on the topic, with a focus on con- tinuous lexical and factual knowledge learning in dialogues, open- domain dialogue learning after deployment and learning of new language expressions via user interactions for language grounding applications (e.g. natural language interfaces). Finally, we conclude with a discussion on the scopes for continual conversational skill learning and present some open challenges for future research.

Target Audience

Researchers, graduate students, and practitioners who are interested in lifelong or continual learning, dialogue systems, and continual learning after deployment. It is particularly suitable for people who have been or will be invovled in building dialogue systems or chatbots.

Prerequisite Knowledge

Basic knowledge of machine learning and natural language processing.