Declarative AI 2023

Abstract: Symbolic or declarative methods have been extensively used in manufacturing, e.g., as digital representations of physical objects or systems, where they have become one of the key building blocks towards digitalization and automation in the whole production value chain. Indeed, rules, ontologies, answer set programs have been used for modelling of industrial assets as Digital Twins, for industrial analytics, integration and querying of production data, process monitoring and equipment diagnostics, moreover, semantic technologies have been adopted or evaluated in a number of large high tech production companies such as Bosch, Equinor, Festo, Siemens, etc. New trends in manufacturing that often referred to as Industry 4.0 and that are characterised by an extensive use of sensors and IoT technology brought enormous volumes of production data from manufacturing facilities. This requires data driven solutions including Machine Learning that can cope industrial big data. At the same time such solutions should account for the declarative representations in order to take the vital manufacturing knowledge captured in them to the full extend thus ensuring trust, reliability, explainability, and transparency of AI solutions. In the keynote talk we discuss the declarative aspects of manufacturing, the new smart manufacturing trends and the necessity of Neural-Symbolic solutions. We will give a number of examples from Bosch and other companies, discuss research and industrial challenges and new exciting directions.

About: Evgeny Kharlamov is a Senior Expert in “AI methods for Semantic Digital Twins and Knowledge Graphs” at the Bosch Center for Artificial Intelligence and an Associate Professor at the University of Oslo. He was previously a Senior Research Fellow at the University of Oxford, a visiting researcher at the University of Edinburgh, and a researcher at the University of Bolzano and INRIA Saclay. Evgeny does AI-centered research that aims at sustainable, circular, and smart manufacturing and centerred around topics of standardised, intelligent and data-driven production value-chain empowered with digital twins and IoT. His research in particular accounts for Semantic Technologies, ontologies, knowledge and cognitive graphs for symbolic representation and reasoning over manufacturing knowledge, for machine learning for processing of production data, and for their combinations in Neural-Symbolic AI methods. Evgeny’s work led to 130+ publications including top tier venues such as NeurIPS, JIM, TODS, PVLDB, SIGMOD, IJCAI, AAAI, CIKM, and ISWC. His citation count at Google Scholar is about 3K. He won several prestigious awards including the best research and industrial applications papers at ESWC’20, ISWC’17, best demo at ISWC’15, and he is ranked as 18th among “AI 2000 Knowledge Engineering Most Influential Scholars” by AMiner.

Abstract: Knowledge Graph Embeddings are representations of entities and relations as vectors in a continuous space, and, as such, are used in many tasks, like link prediction or entity classification. While most evaluations of knowledge graph embeddings are on quantitative benchmarks, it is still not fully understood what they are actually capable of learning. In this talk, I will show how to quantify the representative capabilities of knowledge graph embeddings, and provide insights into what kinds of logical constructs can be represented by which embedding methods. Based on those considerations, I will discuss various ways in which knowledge graph embeddings can benefit from symbolic schema information, and how those combinations open new ways of evaluating knowledge graph embeddings beyond standard metrics.

About: Heiko Paulheim is a Full Professor for Data Science at the University of Mannheim. Prior to joining the University of Mannheim, he worked as a researcher for SAP Research on ontology-based application integration and the Technical University of Darmstadt on machine learning with Linked Open Data. His group in Mannheim conducts various projects around knowledge graphs, yielding, among others, the large-scale public knowledge graphs WebIsALOD, CaLiGraph, and DBkWik. Moreover, his group is concerned with using knowledge graphs in machine learning, which has lead to the development of the widespread RDF2vec method for knowledge graph embeddings. Recent directions also include the embedding of dynamic and of spatio-temporal knowledge graphs. In addition to those topics, Heiko Paulheim also conducts projects which are concerned with ethical, societal, and legal aspects of AI, including KareKoKI, which deals with the impact of price-setting AIs on antitrust legislation, and the ReNewRS project on ethical news recommender systems.

Abstract: The creation and maintenance of knowledge graphs is commonly based on the generation and use of several types of artefacts, including ontologies, declarative mappings and different types of scripts and data processing pipelines, sample queries, APIs, etc. All of these artefacts need to be properly maintained so that knowledge graph creation and maintenance processes are sustainable over time, especially in those cases where the original data sources change frequently. It is not uncommon to have situations where ontologies are governed by an organisation or group of organisations, while mappings and data processing pipelines are handled by other organisations or individuals, using different sets of principles. This causes mismatches in the knowledge graphs that are generated, including the need to update all the associated artefacts (declarative mappings, sample queries, APIs, etc.) so as to keep up to date to changes in the ontologies, or in the underlying data sources. In this talk we will discuss several of the challenges associated to the maintenance of all of these artefacts in real-world knowledge graph scenarios, so as to provide some light into how we could set up a complete knowledge graph governance model that may be used across projects and initiatives.

About: Oscar Corcho is Full Professor at the AI Department at Universidad Politécnica de Madrid (UPM), where he is the deputy director of the R&D Center for Artificial Intelligence (AI.nnovation Space) and academic director of the EC-funded master on Artificial Intelligence for Public Services (AI4Gov), and leads a research group of 40+ members. His core research activities are centered around Open Science, Knowledge-Graph-based Data Integration and Ontological Engineering. These are applied to several domains of expertise, including public and private organisations. Since 2021, he co-leads the EOSC Association Semantic Interoperability Task Force, continuing with the earlier work on the EOSC FAIR working group, where he led the edition of the EOSC Interoperability Framework.

Abstract: . Large language models (LLMs), as a particular instance of generative articial intelligence, have revolutionized natural language processing. In this invited paper, we argue that LLMs are complementary to structured data repositories such as databases or knowledge bases, which use symbolic knowledge representations. Hence, the two ways of knowledge representation will likely continue to co-exist, at least in the near future. We discuss ways that have been explored to make the two approaches work together, and point out opportunities and challenges for their symbiosis.

About: Fabian M. Suchanek is a full professor at the Telecom Paris University in France. He obtained his PhD at the Max-Planck Institute for Informatics under the supervision of Gerhard Weikum. In his thesis, Fabian developed inter alia the knowledge base YAGO, one of the largest public general-purpose knowledge bases, which earned him a honorable mention of the SIGMOD dissertation award. Fabian was a postdoc at Microsoft Research in Silicon Valley (reporting to Rakesh Agrawal) and at INRIA Saclay/France (reporting to Serge Abiteboul). He continued as the leader of the Otto Hahn Research Group “Ontologies” at the Max-Planck Institute for Informatics in Germany. Since 2013, he is an associate professor at Télécom Paris University in France, and since 2016 a full professor. Fabian teaches classes on the Semantic Web, Information Extraction and Knowledge Representation in France, in Germany, and in Senegal. With his students, he works on information extraction, rule mining, ontology matching, and other topics related to large knowledge bases. He has published more than 100 scientific articles, among others at ISWC, VLDB, SIGMOD, WWW, CIKM, ICDE, and SIGIR, and his work has been cited more than 15,000 times. His 2007 paper on YAGO won the Test of Time Award of the WWW 2018 conference.

Abstract: Presenting the results of a multi-year journey applying Declarative AI to a critical government mission.  Leveraging commercial-of-the-self components and innovative design patterns, this journey has combined Deep Neural Network (DNN), and Machine Learning together with Decision Management, and Robotic Process Automation (RPA) to deliver a robotic workforce powered by Declarative AI, performing complex case management alongside human case workers.  Results include substantial gains in efficiency, quality, and consistency verifying the eligibility of tens of millions of consumers seeking to requiring government benefit eligibility verification. One of the largest and most complex applications of AI within the federal government arena, the results of this journey makes a compelling illustration of the difference between Statistical and Declarative AI – at scale!  This session will feature a transparent presentation of our results, metrics, approach and lessons learned (including many never before disclosed to a public audience.)  You will learn how we leveraged Declarative AI to escape and exceed the traditional boundaries of automation, moving from discrete tasks to perform “robo-adjudication” delivering greater accuracy, efficiency, and quality of work. Also demonstrated will be how our Declarative AI is leveraged to assign work to humans and robots, ensuring every time right work performed by the right worked at precisely the right moment.  Nathaniel will show strategies for automation at this massive scale can be executed with full transparency and accountability, eliminating the reliance on subjective interpretation of policies and rules, while delivery more accurate analytics and ensuring program integrity.  He will discuss how to deliver intelligent automation at scale, while avoiding the pitfalls which inevitably otherwise doom to fail initiatives of this size and scope, including what challenges were overcome as well as those unforeseeable at the outset.

About: Previously rated as the #1 Most Influential Thought Leader in Business Process Management by independent research as well as one of the Top 10 Leading Luminaries by Data Informed magazine, Nathaniel Palmer frequently tops the lists of the most recognized names in his field.  He has been featured in media ranging from Fortune to The New York Times, and has and been a guest expert on National Public Radio (NPR).  Nathaniel is a pioneer in the arena of automation and digital transformation, having the led the design for some of the industry’s largest-scale and most complex initiatives, involving investments of $500 Million or more.  He frequently tops the lists of the most recognized names in his field, and was the first individual named as “Laureate in Workflow” as well as a recipient of the Marvin L. Manheim Award For Significant Contributions in the Field of Workflow.  He is a regular speaker at leading forums and industry user groups, and has co-authored over a dozen books on digital transformation including “The X-Economy” (2001), “Intelligent Adaptability” (2017), “BPM Everywhere: Internet of Things and Process of Everything” (2015),  “Passports to Success in BPM” (2014),  “Intelligent BPM” (2013), “How Knowledge Workers Get Things Done” (2012),  “Social BPM” (2011), “Mastering the Unpredictable” (2008) which reached #2 on the Amazon.com Best Seller’s List.  His latest book, “Gigatrends,” to be published in early 2024 defines the leading global trends affecting populations of a 1 Billion or more, each with greater than $1 Trillion in economic impact.