Typing in Context
Steffen Staab, Universität Koblenz-Landau, Germany
Formal concept analysis (FCA) derives a hierarchy of con- cepts in a formal context that relates objects with attributes. This ap- proach is very well aligned with the traditions of Frege, Saussure and Peirce, which relate a signifier (e.g. a word/an attribute) to a mental concept evoked by this word and meant to refer to a specific object in the real world. However, in the practice of natural languages as well as artificial languages (e.g. programming languages), the application context often constitutes a latent variable that influences the interpretation of a signifier. We present some of our current work that analyzes the usage of words in natural language in varying application contexts as well as the usage of variables in programming languages in varying application contexts in order to provide conceptual constraints on these signifiers.
Elements about Hybrid, Exploratory, and Explainable Knowledge Discovery
Amedeo Napoli, Université de Lorraine, CNRS, Inria, LORIA, France
Knowledge discovery (KD) in complex databases and especially the mining of interesting patterns can be read along several dimensions based on data, knowledge, problem-solving and interactivity. These dimensions are not disconnected and have a direct impact on the quality, applicability, and eciency of KD. Accordingly, we discuss four objectives for knowledge discovery which are based on these four dimensions, namely exploration, knowledge orientation, hybridization, and explanation. The data space and the pattern space can be explored in several ways, depending on specic evaluation functions and heuristics, possibly related to domain knowledge. Moreover, knowledge discovery is knowledge oriented and data mining should be car- ried out w.r.t. some domain models that can be used as references. Also, numerical data are complex and supervised numerical machine learning methods are usually the best candidates for eciently mining such data. However, the work and output of numerical methods are most of the time hard to understand. This calls for hybridization, as the combination of numerical and symbolic mining methods may improve the range of applicability and interpretability of knowledge discovery. Finally, suitable explanations about the operating models and possible subsequent decisions should complete any knowledge discovery process, and this is far from being the case at the moment. For illustrating these dimensions and objectives, we analyze a concrete case about the mining of biological data. We characterize these dimensions, their connections and coexistence, and how this can be read within the objectives of knowledge discovery. Finally, we discuss all these elements in the framework of Formal Concept Analysis and try to draw some perspectives for future research.
Too Much Information: Can AI Cope With Modern Knowledge Graphs?
Markus Krötzsch, Technische Universität Dresden, Germany
Knowledge graphs play an important role in artificial intelligence (AI) applications – especially in personal assistants, question answering, and semantic search – and public knowledge bases like Wikidata are widely used in industry and research. However, modern AI includes many different techniques, including ma- chine learning, data mining, natural language processing, which are often not able to use knowledge graphs in their full size and complexity. Feature engineering, sampling, and simplification are needed, and commonly achieved with custom preprocessing code. In this position paper, we argue that a more principled integ- rated approach to this task is possible using declarative methods from knowledge representation and reasoning. In particular, we suggest that modern rule-based systems are a promising platform for computing customised views on knowledge graphs, and for integrating the results of other AI methods back into the overall knowledge model.
Learning Implications from Data and from Queries
Sergei Obiedkov, Higher School of Economics, Moscow, Russia
In this paper, we consider computational problems related to finding implications in an explicitly given formal context or via queries to an oracle. We are concerned with two types of problems: enumerating implications (or association rules) and finding a single implication satis- fying certain conditions. We present complexity results for some of these problems and leave others open. The paper is not meant as a comprehen- sive survey, but rather as a subjective selection of interesting problems.