MAKE stands for Machine leArning and Knowledge Engineering – the challenges that we address during the two-day MAKEathon will benefit from a combination of machine learning and explicit knowledge.
The event will be held physically, registration is mandatory.
The event will take place in the main building of the FHNW University of Applied Sciences And Arts Northwestern Switzerland, Olten (CH). Health protection measures recommended by the Federal Council and the FHNW will be followed. Find the latest news here.
“MAKEathon aims to generate AI-based solutions that address global challenges by combining the strengths of Machine Learning and Knowledge Engineering”
Organizations from both research and industry are more than ever seeking to leverage AI solutions to get a competitive edge. Experience in the practice of Artificial Intelligence has shown that there is no “one-size-fits-all” solution.
Many AI solutions often consider machine learning approaches only. Machine learning helps to solve complex tasks based on real-world data instead of already existing knowledge or pure intuition. It is most suitable for applications in which we are not aware of the knowledge we apply or when knowledge is not known. Think of face recognition, which allows a system to detect and identify people on photos. The larger and the more accurate the training set, the better the outcome. However, if a new scenario (not foreseen in the training set) occurs, the ML-based object detection algorithm may fail by making the wrong classification (e.g. taking the picture of a person on a street advertisement for a real human being). This shows that data-driven solutions are far from being intelligent as they need a huge amount of data to achieve accurate conclusions.
Additionally, there are application areas in which it is important that machines can explain their suggestions. This is particularly the case when decisions can have serious consequences like in banking, insurance, or medicine. In medicine, for instance, a physician might not accept a diagnosis or therapy without adequate explanation. Many businesses are highly regulated and, thus, require compliance with law and regulations. Application-specific domain knowledge can be represented with the help of knowledge-based systems that make knowledge explicit and can explain their conclusions.
Given their complementary strengths and weaknesses, there is increasing attention on AI solutions integrating both approaches – knowledge engineering (KE) and machine learning (ML). During the MAKEathon, we follow this idea to create innovative prototypes and new solution approaches.
MAKEathon includes learning and experts from the AAAI-MAKE Spring Symposium on Combining Machine Learning with Knowledge Engineering. Since March 2019, the AAAI symposium is held at Stanford University on a yearly basis and gathers worldwide AI researchers and practitioners.
Title: Explicit knowledge modeling for AI-driven decision making
Abstract: Artificial Intelligence – be it data-driven or explicit AI – has the potential to revolutionize the way we work. A form of explicit AI, knowledge graphs deliver connected insights across functional or use case boundaries and drive knowledge democratization in the enterprise. Because they support the explicit modeling of knowledge traditionally hidden in complex processes, long documents, domain-specific applications, or domain experts’ minds, they are a great foundation for trustworthy and explainable business decisions and processes.
In this tutorial, we will present the gold standard approach to explicit knowledge modeling (semantic modeling) which transforms the creation of a knowledge graph into a streamlined, end-to-end process where all relevant stakeholders – from IT experts and ontology engineers to domain experts and business users – are equally involved. This approach is based on metaphactory’s visual and user-friendly interface for creating, exploring, visualizing, editing, and documenting knowledge graph assets such as ontologies, taxonomies, or data catalogs. The visual language translates to core elements of OWL, SHACL and SKOS and results in knowledge graph assets based on open and flexible W3C standards.
To make the tutorial as use case driven as possible, we will look at explicit knowledge modeling in the context of skill management and will cover the following topics: