Keynote Speakers

• Sven Dickinson

Sven Dickinson received the B.A.Sc. degree in Systems Design Engineering from the University of Waterloo, in 1983, and the M.S. and Ph.D. degrees in Computer Science from the University of Maryland, in 1988 and 1991, respectively. He is Professor and past Chair of the Department of Computer Science at the University of Toronto, and is also Vice President, Chief Scientist, and Head of the new Samsung Toronto AI Research Center, which opened in May, 2018. Prior to that, he was a faculty member at Rutgers University where he held a joint appointment between the Department of Computer Science and the Rutgers Center for Cognitive Science (RuCCS). His research research interests revolve around the problem of shape perception in computer vision and, more recently, human vision. He has received the National Science Foundation CAREER award, the Government of Ontario Premiere's Research Excellence Award (PREA), and the Lifetime Research Achievement Award from the Canadian Image Processing and Pattern Recognition Society (CIPPRS). He currently serves on eight editorial boards, including the role of Editor-in-Chief of the IEEE Transactions on Pattern Analysis and Machine Intelligence, and the role of co-editor of the Morgan & Claypool Synthesis Lectures on Computer Vision. He is a Fellow of the International Association for Pattern Recognition (IAPR).

The Role of Symmetry in Human and Computer Vision

Symmetry is one of the most ubiquitous regularities in our natural world. For almost 100 years, human vision researchers have studied how the human vision system has evolved to exploit this powerful regularity as a basis for grouping image features and, for almost 50 years, as a basis for how the human vision system might encode the shape of an object. While computer vision is a much younger discipline, the trajectory is similar, with symmetry playing a major role in both perceptual grouping and object representation. After briefly reviewing some of the milestones in symmetry-based perceptual grouping and object representation/recognition in both human and computer vision, I will articulate some of the research challenges. I will then briefly describe some of our recent efforts to address these challenges, including the detection of symmetry in complex imagery and understanding the role of symmetry in human scene perception. In the last part of my talk, I will step back and briefly talk about some of the technical, societal, ethical, and educational challenges we face as an AI community.

• Chad Gaffield

Chad Gaffield is Distinguished University Professor at the University of Ottawa (Canada) where he holds the University Research Chair in Digital Scholarship. An expert on the sociocultural history of Canada, Gaffield has developed computer-based approaches to study socio-demographic change during the 19^th and 20^th centuries, childhood and family history during the initial decades of mass schooling, and the emergence and development of Canada’s official language communities. He has led major interdisciplinary, multi-institutional and cross-sectoral research collaborations, including the building of the Canadian Century Research Infrastructure that developed digital technology to analyze and interpret the temporal and spatial forces that shaped the 20th century. Dr. Gaffield’s awards include the Royal Society of Canada’s (RSC) J.B. Tyrrell Historical Medal and the Antonio Zampolli Prize given by the international Alliance of Digital Humanities Organizations. He was appointed Officer of the Order of Canada in 2017. Dr. Gaffield served as President and CEO of the Social Sciences and Humanities Research Council of Canada from 2006 to 2014, and was elected President of the RSC (2017-2019).

Past, present and possible AI futures:  Will this time be different?

The boom-bust cycle of excitement about AI since the mid-20th century cautions against all-in enthusiasm today, despite encouraging signs that this time will indeed be different. While specialists know this, decisions about research funding and social licensing depend on public discussion that characteristically has swung dramatically between positive and negative assessments. The good news is that AI’s past offers valuable insights to today’s developments while also suggesting promising approaches for continued progress. The bad news is that these insights and promising approaches imply that staying on the current path may well lead to another AI winter if action is not taken. With a view toward candid discussion about AI’s current strengths and weaknesses, this presentation will draw upon the speaker’s experience in historical data creation and analysis as well as his years as President and CEO of the federal granting agency, the Social Sciences and Humanities Research Council of Canada.

• Nancy Ide

Nancy Ide is Professor of Computer Science at Vassar College. She has been an active researcher in the field of computational linguistics for over 30 years and has published copiously on topics including computational lexicography, word sense disambiguation, semantic web implementations of linguistically annotated data, and interoperable standards for representing language resources. In 1987, she co-founded the Text Encoding Initiative, which continues to be the major XML format for representing annotated humanities data; and later developed the XML Corpus Encoding Standard (XCES). She has been Principal Investigator on several National Science Foundation-funded projects, including a major effort to create a linguistically-annotated corpus of American English and an ongoing project to provide a platform including fully interoperable NLP tools and data, the Language Applications (LAPPS) Grid. Dr. Ide is the co-editor-in-chief of the Springer journal Language Resources and Evaluation and editor of the Springer book series entitled Text, Speech, and Language Technology. She is the co-founder and President of the Association for Computational Linguistics special interest group on Annotation (ACL-SIGANN).

Challenges for Scientific Publication Mining

Natural Language Processing (NLP) techniques have been applied to mining scientific publications in order to discover new hypotheses as well as better expose so-called “dark data” that is characteristic of sample-based science. However, despite progress over the past twenty or more years, there remain several stubborn obstacles to efficient and effective exploitation of NLP techniques for scientific publication mining, ranging from a lack of sufficient support for fine-tuning tools to specific sub-fields and/or tasks to inadequate access to the majority of the scientific literature that is produced today. This presentation will survey the challenges for efficient and effective scientific publication mining and describe how Language Applications (LAPPS) Grid currently addresses some of them, for example, by providing facilities for machine-assisted development of gold standard corpora for training domain- and task-specific language models. Finally, I will outline some developing solutions, including the potential to exploit recently developed, state-of-the-art deep learning systems as well as high-powered computing facilities to generate appropriate language models for scientific publications and fine-tune them to specific domains and/or tasks.

• Maite Taboada

Maite Taboada is Professor in the Department of Linguistics at SFU. Her research combines discourse analysis and computational linguistics, with an emphasis on discourse relations and sentiment analysis. Current work focuses on the analysis of online comments, drawing insights from corpus linguistics, computational linguistics and big data. Other projects include a study of fake news online and the Gender Gap Tracker. She is the director of the Discourse Processing Lab at SFU.

Managing comments and misinformation online with text classification techniques

Abusive comments, toxic behaviour, harassment, misinformation and fake news–it seems like a bleak landscape online. Some of these problems can be solved with text classification techniques. If we can identify ‘nice’ and ‘nasty’ comments, then we can promote the former and filter out the latter. If we can reliably identify misinformation and fake news stories, then we can stop them before they spread online.

My lab is actively working on these issues and I will discuss two ongoing research projects. The first project aims at identifying constructive comments on news stories. Using both ‘classic’ machine learning (Support Vector Machines with linguistic features) and deep learning methods, we have built a classifier to identify instances of constructive comments, defined as those that are related to the article, intend to create a civil dialogue and provide specific points supported by evidence. The classifier was built using a large annotated corpus of comments (12,000 comments) from the Canadian daily The Globe and Mail. Our results show that constructiveness can be identified reliably and that a mix of features characterize constructive comments, including length, specific points and the presence of personal stories. The goal of this project is to build a moderation platform to allow constructive comments to be featured more prominently, which will hopefully encourage more constructiveness in online spaces.

The second project addresses the classification of stories into fake and fact-based stories, also using deep learning techniques (CNNs and the ULMFit model). The most challenging part of this project has been the compilation of a dataset with reliably labeled fake news stories. Although one may get the impression that fake news stories are everywhere, it is actually quite challenging to collect a large enough dataset. We cannot simply download stories from certain websites and domains, as those sites may publish both fake and fact-based stories, leading to noisy data. We have instead relied on fact-checking websites such as Snopes and Emergent, which label articles individually. We have so far compiled a dataset of about 10,000 articles, which we are using in our experiments and will make publicly available.