Skip to content

AAAI 2018 Workshop

Natural conversation is a hallmark of intelligent systems. Unsurprisingly, dialog systems have been a key sub-area of AI for decades. Their most recent form, chatbots, which can engage people in natural conversation and are easy to build in software, have been in the news a lot lately. There are many platforms to create dialogs quickly for any domain, based on simple rules. Further, there is a mad rush by companies to release chatbots to show their AI capabilities and gain market valuation. However, beyond basic demonstration, there is little experience in how they can be designed and used for real-world applications that need decision making under constraints (e.g., sequential decision making). The workshop will thus be timely in helping chatbots realize their full potential.
Furthermore, there is an upcoming interest and need for innovation in Human-Technology-Interaction, as addressed in the context of Companion Technology. Here, the aim is to implement technical systems that smartly adapt their functionality to their users’ individual needs and requirements and are even able to solve problems in close co-operation with human users. To this end, they need to enter into a dialog and convincingly explain their suggestions and decision-making behavior.
From research side, statistical and machine learning methods are well entrenched for language understanding and entity detection. However, the wider problem of dialog management is unaddressed with mainstream tools supporting rudimentary rule-based processing. There is an urgent need to highlight the crucial role of reasoning methods such as constraints satisfaction, planning and scheduling, and learning working together with them, that can play to build an end-to-end conversation system that evolves over time. From the practical side, conversation systems need to be designed for working with people in a manner that they can explain their reasoning, convince humans about making choices among alternatives, and stand up to ethical standards demanded in real life settings.
Topics of interest, include, but are not restricted to:
Dialog Systems
  • Early experiences with implemented dialog systems
  • Evaluation of dialog systems, metrics
  • Open domain dialog and chat systems
  • Task-oriented dialogs
  • Style, voice and personality in spoken dialogue and written text
  • Novel Methods for NL Generation for dialogs
  • Domain model acquisition, especially from unstructured text
  • Plan recognition in natural conversation
  • Planning and reasoning in the context of dialog systems
  • Learning to reason
  • Learning for dialog management
  • End2end models for conversation
  • Explaining dialog policy
Practical Considerations
  • Ethical issues with reasoning in dialog systems
  • Corpora, Tools and Methodology for Dialogue System
The intended audience includes students, academic researchers and practitioners with an industrial background from the AI sub-areas of dialog systems, learning, reasoning, planning, HCI, ethics and knowledge representation.
Workshop Agenda
Talk 1:
Title: Continuous Improvement of Intelligent Assistants Through Interaction with Live Customers
Speaker Erik T. Mueller
Abstract: On March 10, 2017, Capital One launched Eno, the first natural language SMS chatbot from a U.S. bank. Eno allows customers to track account balances, see recent transactions, and make credit card payments by texting in natural language. In this talk, I discuss the artificial intelligence capabilities we're building into Eno and how we continuously improve them by leveraging interactions with live customers.
Bio: Erik Mueller is Director of Intelligent Assistants at Capital One. Before joining Capital One, Erik was a member of the original Watson Team at IBM for five years, where he co-developed the Watson Jeopardy! system, Watson for Healthcare, and WatsonPaths. He won the AAAI Feigenbaum Prize with the IBM team. He has ten patents on artificial intelligence and is the author of four books on AI, including Commonsense Reasoning, published by Morgan Kaufmann/Elsevier. He holds a Ph.D. and M.S. in computer science from UCLA and an S.B. in computer science and engineering from MIT.
Talk 2:
Title: Combining Functional Programming, Probabilistic Reasoning, and Machine Learning in an Event-Driven AI Agent Framework
Speaker Matthew (Matt) Davis
Abstract: IBM Research has developed a framework and platform for leveraging probabilistic reasoning and machine learning models in a functional programming framework that allows modularized skills composition from microservices and learning for AI Agents. Our framework leverages both established and novel algorithms for information retrieval, classification, NLP, planning, and event generation and sequencing. We will discuss use cases, implementation, and results from real-world deployments.
Bio: Matthew (Matt) Davis is a Researcher and Manager at IBM Research - Cambridge, USA focusing on customization and generalization of AI Systems.
Talk 3:
Title: Interactions with Social Robots - issues and challenges in combining knowledge and dialogue capabilities for digital companions
Speaker Kristiina Jokinen
Abstract: In recent years, many different robotic companions have appeared to enable potentially useful applications for healthcare assistance, home and assisted living services, language learning, gaming, etc. The role of the robot is a peer-like companion which can provide useful information and communicate with the human users in natural language. Such agents appear as social robots which act and interact in the physical world, and their appearance, behavior, and interaction capabilities affect the users’ views and acceptance of the applications. However, many practical challenges remain to be solved before robot companions become a reality. In this talk I will focus on dialogue modelling that enables interaction between users and social robots. In particular, I will discuss dialogue design that takes into account structured data on human activities so as to expand the system's reasoning and interaction capabilities with the help of goal-directed ontologies. I will also discuss technology for building human-centered interactive systems and present our work to create a framework for automated social agents that can assist human care-takers in their collaborative activities in service industries such as nursing, caregiving, and education. I will also explore the general challenges concerning possibilities to improve and reconstruct operations in service industries based on explicit knowledge and social robots.
Bio: Kristiina Jokinen is Senior Researcher at AI Reserach Center at AIST Tokyo Waterfront. Before joining AIRC, she was Professor and Project Manager at University of Helsinki and at University of Tartu. She received her PhD from UMIST, Manchester, and was awarded the JSPS Fellowship to research at the AIST, Japan. She was Invited Researcher at the ATR Research Labs in Kyoto, and Visiting Professor at Doshisha University in Kyoto in 2009-2010. She was the Nokia Foundation Fellow in Stanford in 2006, and she is Life Member of Clare Hall College at University of Cambridge. Her research focuses on spoken dialogue systems, corpus analysis, and cooperative and multimodal human-robot communication. She has widely published on these topics, including three books. Together with G.Wilcock she developed the WikiTalk open-domain dialogue application for social robots. She has had a leading role in multiple national and international cooperation projects. She has served as General Chair for the SIGDial 2017 and ICMI 2013, Area Chair for Interspeech 2017 and COLING 2014, she organised the northernmost dialogue conference IWSDS in 2016 in Lapland and edited the Springer book "Dialogues with Social Robots" (LNEE 427)
Talk 4:
Title: Statistical Machine Learning for Dialog Management; its history and future promise.
Speaker Paul Crook
Abstract: Partially contradicting the workshop's premise, statistical machine learning (ML) methods for dialogue management have a long history, similar to that of other areas of natural language processing. The talk starts with a  review of the statistical approaches that have been applied in the literature, especially to goal-orientated dialogues such as browsing and task-completion. Techniques such as modelling dialogues as sequential decision-making processes, e.g. MDPs and POMDPs. Then suggest reasons why such approaches have struggled to cross over into industry.
The second part of the talk  considers the new golden age that has opening-up for dialogue management. The recent influx of Deep Learning approaches that remove the burden of input featurization and dialogue-state design. Sequence-to-Sequence and Information Retrieval methods that make it easy to stand-up shallow chatbots that despite their lack of understanding are able to converse in convincingly natural language. Search Engines that are attempting to incorporate elements of dialogue state-tracking and deliver actions, e.g. deep links to Apps, to users. The confluence of these approaches offers the potential for exciting new ways of building and training dialogue systems, repacking statistical dialogue management in forms that non-experts can use, and thus deliver on some of the promises of Conversational AI.
Bio: Paul Crook is currently a Senior Scientist at Microsoft where over the past five years he has contributed to the design and implementation of the conversational engine of Cortana; introducing the concepts of Information-State (IS) dialogue management and statistical ranking of dialogue hypotheses. Prior to Microsoft, Paul was a Research Fellow in the Interaction Lab at Heriot-Watt University and Research Associate at the University of Edinburgh. His research focuses on statistical machine learning methods for dialogue management and user simulation, especially Reinforcement Learning and more recently Deep Learning.
Accepted Papers (Full Presentation)
  • MEETING BOT: Reinforcement Learning for Dialogue Based Meeting Scheduling  Vishwanath D, Lovekesh Vig, Gautam Shroff and Puneet Agarwal
  • Knowledge-Graph Driven Information State Approach to Dialog Svetlana Stoyanchev and Michael Johnston
  • An Inherently Explainable Model for Video Activity Interpretation Sathyanarayanan Aakur, Fillipe de Souza and Sudeep Sarkar
  • Building Dialogue Structure from Discourse Tree of a Query Boris GalitskyProduction Ready Chatbots: Generate if not Retrieve Aniruddha Tammewar, Monik Pamecha, Chirag Jain, Apurva Nagvenkar and Krupal Modi
  • Specifying and Implementing Multi-Party Conversation Rules with Finite-State-Automata Maira Gatti de Bayser, Melina Alberio Guerra, Paulo Cavalin and Claudio Pinhanez
  • State Tracking Networks For Dialog State Tracking Xuguang Wang, Xingyi Cheng, Jie Zhou and Wei Xu
Accepted Papers (Poster)
  • Gated Orthogonal Recurrent Units: On Learning to Forget Li Jing, Caglar Gulcehre, John Peurifoy, Yichen Shen, Max Tegmark, Marin Soljacic and Yoshua Bengio
  • `Did I Say Something Wrong?'': Towards a Safe Collaborative Chatbot Merav Chkroun and Amos Azaria
  • Making Personalized Recommendation through Conversation: Architecture Design and Recommendation Methods Sunhwan Lee, Robert Moore, Guang-Jie Ren, Raphael Arar and Shun Jiang
  • Automatic Extraction of Domain Specific Latent Beliefs in Customer Complaints to help tailor Chatbots Amit Sangroya, C. Anantaram, Pratik Saini and Mrinal Rawat
  • Context Dependent Additive Recurrent Neural Network for Dialog Systems Quan Hung Tran, Trung Bui and Hung Bui
Important Dates
Manuscripts due:                        November 9, 2017
Notification of acceptance:     November 19, 2017
Camera-ready manuscript:     November 21, 2017
Workshop:                                      February 2, 2018
Submission Guidelines 
Papers must be formatted in AAAI two-column, camera-ready style (AAAI style files are at: ).
Regular research papers, which present a significant contribution, may be no longer than 7 pages, where page 7 must contain only references, and no other text whatsoever.
Short papers, which describe a position on the topic of the workshop or a demonstration/tool, may be no longer than 4 pages, references included.

Program Committee

  • Adi Botea, IBM
  • Kartik Talamadupula, IBM
  • Walter Lasecki, University of Michigan
  • Vera Liao, IBM
  • Pavan Kapanipathi, IBM
  • Mausam, IIT Delhi
  • Dinesh Raghu, IIT Delhi and IBM
  • Sachindra Joshi, IBM