Julia Taylor, Purdue University, USA
What Can Ontology and Natural Language Do For Each Other?
A true ontology should provide a world model for the described domain, identifying its main concepts and tying them together with all relevant contentful properties. The question is how to get this model from text accurately? Assuming, as we do, that there is an accurate and unambiguous way of getting explicitly stated information from text, a lot of information is, in fact, implicit and yet crucial to the world model that we are creating. This implicit information has to be made explicit at the reasoning stage if we hope to come up with the results similar to human reasoning or understanding. And then again, some words in the text are unknown to a machine. Ontology should provide enough knowledge to guess the meaning of these words, as robustly as people do. In this talk, we will look at various ways, requiring optimal human-computer hybrid collaboration, in which ontology helps text understanding and text helps with dynamic ontology development.
Julia M. Taylor earned her Ph.D. in Computer Science and Engineering in 2008, M.S. in Computer Science in 2004, and B.S. in Computer Science and B. A. in Mathematics in 1999 at the University of Cincinnati. After a short postdoc at the Cincinnati Children Hospital and Medical Center, she worked as Senior Research Engineer at RiverGlass, Inc. (2008-2011), improving and implementing the Ontological Semantic Technology for an analytics application. She was a Visiting Scholar at Purdue since 2008 and joined Center for Education and Research for Information Assurance and Security (CERIAS) as a Research Assistant Professor in June 2011. Since January 2012, she has been a tenure-track Assistant Professor of Computer and Information Technology as well as an elected Fellow of CERIAS. She has published a number of peer-reviewed papers in artificial intelligence, computational humor, computational semantics, fuzzy logic, information security, and robotic communication and intelligence. She has also served on the editorial boards of HUMOR: International Journal of Humor Research and European Journal of Humour Research.
Stefan Dietze, L3S Research Center, Germany
From Data to Knowledge – Profiling and Interlinking Datasets on the Web
While the Web of Data, and in particular Linked Data, has seen tremendous growth over the past years, take-up, usage and reuse of data is still limited and is often focused on well-known reference datasets such as DBpedia, Yago or Freebase. Datasets vary heavily with respect to their characteristics, such as the amount, quality, domain coverage or currency of exposed data. While this heterogeneity is natural, given the distributed nature of Linked Data, it also poses challenges for finding, adopting and reusing data across the Web. This problem is further elevated by the lack of reliable information (or metadata) about such dataset characteristics. Hence, judging the suitability and trustworthyness of arbitrary datasets for a given task is challenging. In addition, the evolving nature of Linked Data calls for scalable and efficient techniques which are able to automatically (re-)compute dataset metadata (profiles) and links. This talk will provide an overview of challenges and ongoing research on dataset profiling and interlinking. Profiling involves the automated extraction of dataset metadata while interlinking investigates methods for linking not only entities across the Web of data but also to identify candidate datasets for interlinking tasks. The talk will introduce background, methods, and resulting datasets aimed towards the overall goal of improving take-up and reuse of Web datasets..
Dr. Stefan Dietze is a Research Group Leader at the L3S Research Center (Germany) which he joined in 2011 following previous positions at the Knowledge Media Institute of The Open University (UK) and the Fraunhofer Institute for Software and Systems Engineering (Germany). His research interests are in Semantic Web and Linked Data technologies and their application to Web data integration problems in actual application domains. Stefan currently is leading the European R&D projects LinkedUp (http://linkedup-project.eu) and DURAARK (http://duraark.eu), both dealing with advancing Linked Data technologies and their take-up in real-world application settings. He also is involved in a number of other projects and initiatives in the Semantic Web area and is co-chair of the KEYSTONE (http://www.keystone-cost.eu/) working group on “Representation of Structured Data Sources”. His work has been published in numerous conferences and journals, he is member of many organisation and programme committees and editorial boards and a frequent invited speaker.
Vladimir Gorodetsky, St. Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences
Ontology and User Profiling in 3G Recommender Systems
3G recommender system (3G RecSys) is a novel and, in basic aspects, forthcoming paradigm of human decision support/prediction system. It is expected that 3G RecSys will operate based on human-like semantic categories in knowledge representation, knowledge discovery and knowledge usage. According to the experts’ opinion, 3G RecSys will focus on semantically transparent personal user profile structuring his/her multidimensional personal interests and preferences. 3G RecSys will bring some novel perspectives from data mining and knowledge discovery, while emphasizing user’s decision explanation as well as detection of context-aware causality and incentives determining this or that user’s choice. As an effect of these capabilities, 3G RecSys should be capable, in what concerns with recommendation proposed by it, to answer “why?“ questions using explanatory user interface.
At present days, ontology is recognized as the most natural well-developed modeling framework that is capable to support for 3G RecSys operation with semantic categories specifically tuned for human-like knowledge-based reasoning what is the main expected property of user’s interest representation and its usage in recommendation procedure. Although ontology-based user profile model is being developed from recent times only, to present days, it has become about mature approach.
The talk will present a critical survey on current state-of-the-art with ontology-based user profiling for 3G RecSys. In fact, the reviewed publications prepared the modern viewpoint on the role of ontology as a unified conceptual knowledge modeling framework and knowledge representation structure for 3G RecSys. In the talk, after short introduction into RecSys and into existing user profiling and decision making approaches, an overview of the earlier and contemporary research on ontology-based used profile modeling will be presented.
Next, the novel problems of ontology design inspired by specific data sources available for user profile learning will be highlighted. One of such data sources is the set of tags assigned by users to the web information concerning with particular products or services. Tag-based information is capable to provide for semantic enrichment of the recommendation domain ontology and, consequently, user profile. The talk analyses some aspect of tag-based automated ontology categorization for user profiling using wikification and folksonomies. One more such novel information source that will be outlined in the talk is social network context.
In conclusion, some perspectives for ontology-based user profile learning using both Formal Concept Analysis and Causality Analysis, as well as the author’s experience in this respect are outlined.
Professor of Computer Science, Senior Researcher in Intelligent Systems Laboratory of the St. Petersburg Institute for Informatics and Automation of the Russian Academy of Science. He graduated from the Military Air Force Engineer Academy in St. Petersburg (1960) and Mathematical and Mechanical Department of the St. Petersburg State University (1970), received his Ph.D. degree (1967) and Doctor of Technical Sciences degree (1973) in the area “Space Vehicle Optimal Control”. Main publications (over 250) are related to the areas of multi-agent systems, optimal control system theory, space mechanics, applied statistics, planning, pattern recognition and artificial intelligence, knowledge discovery from databases, data and information fusion, digital image steganography, and computer network security.