NLP for Conversational AI
Co-located with EMNLP 2021
Since the dawn of Artificial Intelligence (AI) research, mathematicians, linguists, and computer scientists have dedicated their careers to empowering human-machine communication in natural language, arguably one of the hardest tasks for an AI. While in recent years the emergence of virtual personal assistants such as Siri, Alexa, Google Assistant, and Cortana has pushed the field forward, the development of such conversational agents remains difficult with numerous unanswered questions and challenges.
Following the success of the 1st and 2nd NLP for Conversational AI workshops at ACL, "The 3rd NLP for Conversational AI” will be a one-day workshop including keynotes, posters, and panel sessions. In keynote talks, senior technical leaders from industry and academia will share insights on the latest developments of the field. An open call for papers will be announced to encourage researchers and students to share their prospects and latest discoveries. There will also be a panel discussion with noted conversational AI leaders focused on the state of the field, future directions and open problems across academia and industry.
Sound exciting? We are looking forward to seeing you!
August 10th, 2021August 15th, 2021 (23:59 GMT-12)
Fast-Track Paper Submission (forwarding EMNLP reviews):
August 28 th, 2021August 29th, 2021 (23:59 GMT-12)
Note: To qualify as a fast-track submission, the EMNLP 2021 reviews need to be uploaded into the system as supplement materials.
Notification of Regular Paper Acceptance: September 17th, 2021
Notification of Cross-Submission Acceptance: September 22nd, 2021
Camera-Ready Paper Due: September 24th, 2021
Workshop Date: November 10th, 2021
Koichiro Yoshino, Institute of Physical and Chemical Research (RIKEN) Robotics Project
Keynote: Embodied dialogue by autonomous robots
The time when robots will support us in our daily lives is just around the corner. There are many problems when we try to build robots that can interact with humans in their living space and assist them in their daily lives. One of the major problems is the robot's perception in the real world. The robot has to recognize objects and actions using various sensors and consider the uncertainty caused by first-person sensors. The robot's goal is also a problem. What the robot should recognize depends on its goal and purpose. The robot should decide its actions according to them. The last issue is about the explainability of the robot; the robot needs to be able to correctly explain why it performed the action and what action is actually performed. We will introduce our works to tackle these problems.
Keynote: Ethical and Technological Challenges of Conversational AI
Conversational AI (ConvAI) systems have applications ranging from personal assistance, health assistance to customer services. They have been in place since the first call centre agent went live in the late 1990s. More recently, smart speakers and smartphones are powered with conversational AI with similar architecture as those from the 90s. On the other hand, research on ConvAI systems has made leaps and bounds in recent years with sequence-to-sequence, generation-based models. Thanks to the advent of large-scale pre-trained language models, state-of-the-art ConvAI systems can generate surprisingly human-like responses to user queries in open domain conversations, known as chit-chat. However, these generation-based ConvAI systems are difficult to control and can lead to inappropriate, biased and sometimes even toxic responses. In addition, unlike previous modular conversational AI systems, it is also challenging to incorporate external knowledge into these models for task-oriented dialog scenarios such as personal assistance and customer services and to maintain consistency.
With great power comes great responsibility. We must address the many ethical and technical challenges of generation-based conversational AI systems to control for bias and safety, consistency, style, knowledge incorporation, etc. In this talk, I will introduce state-of-the-art generation-based conversational AI approaches and will point out the remaining challenges of conversational AI and possible directions for future research, including how to mitigate inappropriate responses. I will also present some ethical guidelines that conversational AI systems can follow.
Mona Diab, George Washington University & Facebook
Keynote: Toward Trustworthy Evaluation Frameworks for NLG
In this talk I describe trustworthy evaluation frameworks for NLG specifically for MT and Abstractive summarization emphasizing faithfulness as a desirable metric to optimize for. I contend that Faithfulness correlates well with notions of practical accuracy. I then expand on how can we consider faithfulness for conversational AI systems.
Idan Szpektor, Google
Keynote: Challenges in Informational Dialogues
In this talk, I will focus on informational dialogues, such as domain exploration and recommendation dialogues. I will discuss some of their challenges and present recent research that aims to address these challenges, including automatic evaluation of factual consistency in grounded generative models, retrieval of content in open-ended dialogues and response planning with RL on utterance-based action space.
Tsung-Hsien Wen, PolyAI
Keynote: Grounding Conversational AI research: from sigmoid functions to talking machines
From sigmoid functions to talking machines, we have gone a long way. In this talk, I want to share PolyAI’s journey on productionising cutting-edge conversational AI research, as well as our latest efforts to make machines even more intelligent and understanding. Looking back, we should celebrate the achievements we have made as a community. Looking forward, we need to be aware of the challenges ahead of us.