Knowledge-aware Zero-shot Learning (K-ZSL): Concepts, Methods and Resources
Duration: Half Day
Organizers: Jiaoyan Chen, Yuxia Geng, Yufeng Huang and Huajun Chen
Abstract: Zero-shot Learning (ZSL), which enables machine learning models to predict new targets without seeing their training samples, has attracted wide research interests in many communities such as computer vision (CV) and natural language processing (NLP). An effective solution is to use external knowledge such as text, attribute descriptions and Knowledge Graphs (KGs) to bridge the gap between the targets with training samples and the targets without training samples. This tutorial aims to introduce ZSL from the perspective of knowledge especially KG. We will first present the background of KG and ZSL, then introduce the overall picture of KG-aware ZSL and some representative paradigms with case studies, and finally provide a hands-on session on benchmarks and codes.
Tutorial on Semantic Schema Discovery: Principles, Methods and Future Research Directions
Duration: Half Day
Organizers: Kenza Kellou-Menouer, Nikolaos Kardoulakis, Georgia Troullinou, Zoubida Kedad, Dimitris Plexousakis and Haridimos Kondylakis
Abstract: The explosion of the data on the semantic web has led to many weakly structured, and irregular data sources, becoming available every day. The schema of these sources is useful for a number of tasks, such as source selection, query answering, exploration and summariza- tion. However, although semantic web data might contain schema in- formation, in many cases this is completely missing or partially defined. Schema discovery consists in extracting schema-related information from the original semantic graph, which some applications can exploit instead of or along with the original graph, to perform some tasks more efficiently. This tutorial presents a structured analysis and comparison of existing works in the area of semantic schema discovery helping researchers and practitioners to understand the challenges in the area; it is based upon a recent survey we authored.
A Beginner’s Guide to Reasoning: How to reason your way to better data
Duration: Half Day
Organizers: Valerio Cocchi
Abstract: Reasoning has become an increasingly valued tool in the semantic web space, and yet to many it’s still a black box solution. Perhaps more tragically, despite the explosion of its development in recent years, many in the space still perceive it as a slow, cumbersome, and ultimately impractical technology, which is far from true today. Whether you’re looking to harness reasoning for your own goals, or to peek behind the curtains of someone else’s solution, now is your time to learn. Get hands on with a reasoning engine in this interactive walkthrough: A Beginner’s Guide to Reasoning. You’ll come away understanding the power of reasoning, what it can add to your data, and the fundamentals of how to apply it yourself. With technology in this space running away, there’s never been a better time to learn! This tutorial will touch on the basics of SPARQL, OWL, and Datalog, before diving into reasoning at a technical level. Each participant will come away having built a reasoning solution for themselves, guided along the way by knowledge engineers and subject experts. No prior knowledge is required.
Knowledge-infused Learning for Autonomous Driving (KL4AD)
Duration: Half Day
Organizers: Ruwan Wickramarachchi, Cory Henson, Sebastian Monka, Daria Stepanova and Amit Sheth
Abstract: Autonomous Driving (AD) is considered as a testbed for tackling many hard AI problems. Despite the recent advancements in the field, AD is still far from achieving full autonomy due to core technical problems inherent in AD. The emerging field of neuro-symbolic AI and the methods for knowledge-infused learning are showing exciting ways of leveraging external knowledge within machine/deep learning solutions, with the potential benefits for interpretability, explainability, robustness, and transferability. In this tutorial, we will examine the use of knowledge-infused learning for three core state-of-the-art technical achievements within the AD domain. With a collaborative team from both academia and industry, we will demonstrate recent innovations using real-world datasets.
Artificial Intelligence Techniques for Earth Observation Data (AI4EO)
Duration: Full Day
Organizers: Manolis Koubarakis, Begüm Demir, Dimitris Bilidas, Theofilos Ioannidis, Despina-Athanasia Pantazi, Dharmen Punjani, George Papadakis, George Stamoulis and Eleni Tsalapati
Abstract: In the last few years, there has been a lot of research in applying Artificial Intelligence techniques to Earth observation data. The subareas of Artificial Intelligence that contributed the most to the science of satellite data. are Deep Learning and Semantic Technologies (Ontologies and Linked Data). This tutorial will survey the latest state of the art in this area. It will start by explaining what satellite data is and why satellite data is a paradigmatic case of big spatiotemporal data amenable to Artificial Intelligence techniques. Examples of big satellite data, information and knowledge will be given for the case of the Copernicus program of the European Union. We will teach the tutorial participants how to “break satellite data silos open” by publishing the metadata of satellite datasets as microformats to enable their discovery by modern search engines through services like Dataset Search of Google, how to extract important geospatial information from satellite datasets using deep learning technologies, how to interlink this information with other relevant information available on the Web, and how to make this wealth of data and information freely available on the Web to enable the easy development of geospatial applications. We will present a complete pipeline that starts with satellite datasets in various formats that are made freely available in the archives of space agencies and ends with the deployment of an interactive visual application that uses satellite data utilizing linked data technologies. We will also present a query answering system over geospatial knowledge graphs, that allows non-experts to access linked geospatial data using natural language. The tutorial will give an in-depth coverage of the relevant techniques, systems and some applications developed by the presenters in the last 12 years in the context of 1 ERC grant (BigEarth), 12 European projects (TELEIOS, LEO, Melodies, Optique, BigDataEurope, Copernicus App Lab, WDAQUA, ExtremeEarth, AI4Copernicus and DeepCube), 1 ESA project (Prod-Trees), 3 projects funded by the German government (BIFOLD, TreeSatAI, IDEAL-VGI) and 2 projects funded by the Greek government (SCARE and GeoQA). The two teams presenting the tutorial (from the National and Kapodistrian University of Athens and the Technische Universität Berlin) come from different disciplines (Computer Science and Satellite Remote Sensing) and will offer an interdisciplinary presentation of the relevant theoretical and practical issues.