RuleML 2014

The 8th International Web Rule Symposium
Prague, Czech Republic, August 18-20, 2014

Keynotes and Tutorials

Complex Event Processing (CEP) is an enabling technology to extract actionable, situated knowledge from large amounts of event data in real-time. The promises of the combination of event processing and semantic technologies is that these Semantic CEP (SCEP) engines can use semantic background knowledge for defining more expressive event detection patterns, for understanding what is happening in terms of events and situations, and for knowing what rule-based actions they can invoke. The challenge lies in the combination of distributed real-time big data processing and semantic reasoning with background knowledge bases. The keynote follows the functional layers for preparation, analysis, detection and reaction of the Reference Architecture from the Event Processing Technical Society (EPTS-RA). It addresses the problem of optimized semantic querying, fusion and enrichment of real-time event streams with background knowledge and the expressive reasoning with such enriched events in the higher layers of rule-based event processing and reaction functions. It reports on the standardization activities in Reaction RuleML as a platform-independent rule-based Event Processing Language (EPL) to support such SCEP functions in distributed rule-based Event Processing Networks (EPN) and Event Processing Agents (EPA).

Bio: Prof. Dr. Adrian Paschke is head of the Corporate Semantic Web group (AG-CSW) at the institute of computer science, department of mathematics and computer science at Freie Universitaet Berlin (FUB). He is active in standardization of rule-based event processing as co-chair of the Reaction RuleML technical group, co-chair of the Event Processing Technical Society (EPTS) Reference Architecture working group, and co-editor of the W3C Semantic Web Rule Interchange Format (RIF) standard. He was involved in several research projects such as the EU Network of Excellence "Reasoning on the Web with Rules and Semantics" (REWERSE, 2004-2008) and he currently coordinates and leads several nationally and internationally funded research projects such as the Corporate Semantic Web project (CSW, 2008-2013), the Transatlantic BPM Education Network (BPM EduNet, 2010-2013), the Corporate Smart Content project (CSC, 2013-2016) and the Pragmatic Web community. He is the project leader of several successful open-source projects such as the Prova rule engine, the Rule Responder Pragmatic Web Agent Architecture and the Rule-based Service Level Agreements (RBSLA) project. He has co-authored over 100 papers in the area of Semantic Web, Web Rule Technologies and Semantic Complex Event Processing. He organized conferences such as the ACM Distributed Event Based Systems (DEBS 2012), the RuleML Web Rule Symposium, AIS SigPrag ACM International Conference on Pragmatic Web and the Event-Driven Business Process Management (edBPM) workshops series.

Rules represent knowledge about the world that can be used for reasoning. However, the world is inherently uncertain, which may affect both rules and data. Indeed, rules capturing expert knowledge are only an approximation of a complex reality, and data may be uncertain due to missing values, noisy measurents, or ambiguities.
While a wide variety of formalisms and techniques exist to cope with uncertainty, the approach taken will be based on probabilistic (logic) programming. More specifically, it shall be centered around the prob- abilistic Prolog, ProbLog (see also http://dtai.cs.kuleuven.be/ problog/), which extends the programming language Prolog with probabilistic facts and is based on Sato’s distribution semantics. It combines the deductive power of Prolog with the ability to state the belief that certain facts are true, very much as in probabilistic databases. As such it is a natural rule-based representation for dealing with uncertainty. ProbLog supports probabilistic inference, that is, it can compute the probability P(Q|E) of a query Q given some evidence E.
It also supports learning. To learn parameters, it starts from examples that are partial interpretations (that is, partial descriptions of a possible world), and employs an Expectation-Maximisation approach. ProbLog rules can be learned using a generalization of traditional rule-learning algorithms. These rules are learned form uncertain data. ProbLog has been applied to a number of applications in domains such as bioinformatics, action- and activity recognition and robotics.

Bio: Luc De Raedt is a full professor (of research) at the University of Leuven (KU Leuven) in the Department of Computer Science and a former chair of Machine Learning at the Albert-Ludwigs-University in Freiburg. Luc De Raedt has been working in the areas of artificial intelligence and computer science, especially on computational logic, machine learning and data mining, probabilistic reasoning and constraint programming and their applications in bio- and chemoinformatics, vision and robotics, natural language processing, and engineering. His work has typically crossed boundaries between different research areas, often working towards an integration of their principles. He is well-known for his early work on inductive logic programming (combining logic with learning). Since 2000, he has been working towards a further integration of logical and relational learning with probabilistic reasoning (statistical relational learning and probabilistic programming, on inductive querying in databases, and on combining constraint programming principles with data mining and machine learning. He is currently coordinating a European IST FET project in this area (ICON — Inductive Constraint Programming) and was program (co)-chair of ECAI 2012, ICML 2005 and ECML/PKDD 2001.

The induction of logical rules is among the oldest techniques in machine learning and data mining. Rule learning techniques have been proposed for propositional and relational domains, for predictive and descriptive learning problems, or for supervised or unsupervised tasks. In this tutorial, we will discuss fundamental concepts such as learning algorithms, heuristic functions, and pruning techniques that are used in traditional rule learning algorithms such as Foil, CN2 or Ripper. These will be illustrated in a scenario of learning of a concept given positive and negative traning examples, and we will see how this basic setting can be used for tackling more complex problems such as preference learning, ranking or multilabel classification.

Bio: Johannes Fürnkranz obtained Master Degrees from the Technical University of Vienna and the University of Chicago, and a Ph.D. from the Technical University of Vienna with a Thesis on "Pruning Algorithms for Rule Learning". Since January 2004, he is Professor for Knowledge Engineering at the TU Darmstadt. His main research interest is machine learning, in particular inductive rule learning and preference learning, and their applications in game playing, web mining, and data mining in the social sciences. He is action editor of the journals "Machine Learning" and "Data Mining and Knowledge Discovery" and a regular member of program committees of premier conferences in the areas of machine learning, data mining, information retrieval, and artificial intelligence. He served as the co-chair of the 6th European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (Berlin, 2006), of the 27th International Conference on Machine Learning (Haifa, 2010), and of the 16th International Conference on Discovery Science (Singapore, 2013).

Physicians and medical staff have to come to decisions very often in a complex situation with a huge amount of information available - but not accessible - in a very short time frame.This information is very often spread in different applications, departments, and media. This knowledge is a core asset of a hospital. So centralising this knowledge, storing it in a structured way, versioning it and making it available in these complex situations in an executable way is the functionality of CKMS (clinical knowledge management system) developed by Semedy. Besides storing structured entities CKMS also stores more complex knowledge in rule form. Semedy provides a logic rule based inference engine for that purpose, which combines ontologies, rules and big data. In this talk use cases for CKMS, the architecture of CKMS and of the rule engine are provided. On top medical solutions of this system are sketched.

Bio: Prof. Dr. Jürgen Angele is currently head of development at Procitec GmbH and CTO, director, and shareholder of Semedy AG. Semedy has been cofounded by him in 2011. Semedy develops clinical decision support systems. Formerly he was head of development and shareholder of Ontoprise GmbH, a provider of semantic technologies. Ontoprise has been cofounded by him in 1999. In 1994 he became a full professor in applied computer science at the University of Applied Sciences, Braunschweig, Gemany. From 1989 to 1994 he was a research and teaching assistant at the University of Karlsruhe, Institute AIFB. He did research on the execution of the knowledge acquisition and representation language KARL, which led to a Ph.D. (Dr. rer. pol.) from the University of Karlsruhe in 1993. From 1985 to 1989 he worked for the companies AEG, Konstanz, Germany, and SEMA GROUP, Ulm, Germany. He received the diploma degree in computer science in 1985 from the University of Karlsruhe. He published more than 100 papers as books and journal, book, conference, and workshop contributions. Topics were about semantic web, semantic technologies, knowledge representation, and their practical applications. He was heading a lot of public funded and commercial projects. He gave more than 55 courses at Berufsakademien, Fachhochschulen and Universities. Topics were about: Expert Systems, Software Engineering, World Wide Web, Database Systems, Digital Image Analysis, Computer Graphics, Mathematics.

The idea behind ontology-based data access (OBDA) is to enable users to access data using their own vocubularies and conceptualizations, irrespective of how the data is stored. While users’ conceptualizations are captured in an ontology, the concepts of the ontology are linked to the data sources through declarative mappings. A query q posed in the vocabulary of the ontology is rewritten in a two-stage process: First, q is expanded into a query q’ that encodes relevant parts of the ontology; second, q’ is unfolded into a query q’’ that can be executed on the source systems. In going from q to q’, only information in the ontology is exploited, while; in going from q’ to q’’ only information in the mappings is used. The mappings can be viewed as a complex set of rules formulated in a restricted subset of first-order logic. In this talk we will take a close look at how key notions related to consistency and optimization are to be understood in the context of such rule systems, and review some of their basic properties. We will situate the rule system more closely within the query rewriting process and illustrate particular sources of complexity. Finally we will visit some open problems in this field. OBDA is now being tested on large industrial use cases in the Optique project (www.optique-project.eu), from which example material for this talk is taken.

Bio: Arild Waaler is professor in logic and compuation at the University of Oslo, where he leads the ”Logic and Intelligent Data” research group. He is also the leader of the Optique project, a 14 million Euro FP7 project that runs until the end of 2016. He research interests include automated reasoning, non-monotonic reasoning and proof theory.