Review of the 10th International Semantic Web Conference R. 1 1 School of Computing University of Leeds Intelligent Machines in Synergy with Humans
Outline 1 Conference Overview 2
Conference Overview October 23-27 2011, Bonn, Germany Workshops, Tutorials, other events Main conference: research track, in-use track
Conference Workshops geographical data, ontology matching, uncertainty reasoning, scalable KB systems, ordering and reasoning, knowledge evolution, social data, personalized information management, liked data, linked science, multilinguality, sensors, software engineering, knowledge extraction
Other events Tutorials on e-commerce, processing linked data, debugging ontologies, sparql, user studies, IBM watson, SW app development Posters and Demos with Minute Madness, SW challenge, Linked Data-a-thon, Panel: SW death match Industry vs Academia vs Standards
Main conference: Research Track RDF queries: alternative approaches, performance issues, multiple sources RDF data analysis, Web of Data KR reasoners, semantics, Formal Ontology & Patterns, Justifications & Provenance Ontology evaluation, Ontology Matching Social web Policies & Trust
Main conference: In-use (Industry) Track architecture ontologies and data environmental data content management applications
Repairing ontologies for incomplete reasoners Giorgios Stoilos et al., Oxford Focus on OWL2 RL (e.g. OWLim, Jena, Oracle): Highly scalable but only a subset of OWL, so questions will miss answers. Try to improve completeness while keeping performance through materialisation, inference of axioms in pre-processing stage. Abstraction of reasoners by notion of reasoning algorithm Generate ontology repairs for subset of GALEN ontology to answer LUBM test queries.
QueryPIE: Backward reasoning for OWL Horst over very large knowledge bases Jacopo Urbani et al., VU Amsterdam Improve reasoning performance over large knowledge bases Focus on OWL Horst (aka pd* ruleset) Combine forward and backward reasoning to get best of both worlds forward reasoning off-line on T-Box (schema) terminology-independent reasoning: backward reasoning at query-time only on A-Box (T-Box is already cached) evaluation on LUBM (artificial dataset), LinkedLifeData and FactForge datasets (1 to 10 fold improvement)
Concurrent classification of EL ontologies Yevgeny Kazakov et al., Oxford Distributed implementation of saturation algorithm for OWL EL. to distribute reasoning, axioms in ontology are assigned to contexts Evaluation using SNOMED, GALEN, FMA and GO and comparing with tableaux-based reasoners (Pellet, FaCT++) and saturation-based reasoners without multi-core support (CB, jcel, Snorocket). Up to 2.6 speedup compared to best saturation-based reasoner using 4 workers (tableaux-reasoners much slower on EL ontologies).
An ontology design pattern for referential qualities Jens Ortmann et al., Múnster Model the quality of an entity in referece to another entity E.g. vulnerability of x to factory y Proposed design pattern based on Kuhn s Semantic Reference Systems, implemented on top of DOLCE and used to model vulnerability, resilience and affordances in an ecology domain. thorough (?) analysis of impact of design pattern on use cases, existing design patterns, logical inferences, etc.
Strukt a pattern system for integrating individual and organizational knowledge work Ansgar Scherp et al., Koblenz-Landau Knowledge at organisational level: contracts, orders, etc. can be captured in structured workflows, business process management Knowledge at individual level is harder to capture due to complexity and variability but is present in documents, drafts, calendars, etc. proposal: individual level as weakly structred workflows that capture: descriptions (roles of individuals in the organisation) and situations (goals, events, actions) strukt core ontology (also define higher level patterns such as conditions, resources, status, scheduling) prototype to show how ontology can be used to encode workflows
Encyclopedic knowledge patterns from wikipedia links Aldo Gangemi, Valentina Presutti. STLab, Rome and Bologna how to organise knowledge so that it is easy to grasp? Which should be the base concepts to use? These questions also relevant when building encyclopedias. Hypothesis: structure of Wikipedia (links and resource types) can be used to infer base concepts and existing patterns. Introduces several indicators based on Wikipedia (and DBPedia) structure such as number of resources that have a type, number of times a resource is a subject, path popularity, number of distinct paths, etc. defines an Encyclopedic Knowledge Pattern in terms of the introduced indicators (based on a threshold) User evaluation to determine users agreement
Watermarking for ontologies Fabian M. Suchanek et al., INRIA and Bourgogne Prove that a knowledge base has been copied without permission Current approach is to introduce incorrect facts. Proposal: remove a small percentage of the facts. Evaluation calculating number of facts that need to be removed to have more or less confidence on detecting copied KB
The cognitive complexity of OWL justifications Matthew Horridge et al. Manchester Building on ongoing work on justifications (precise, laconic) Proposes a cognitive complexity model for justifications in order to predict how hard it will be for somebody to understand a justification. Model uses weighted indicators such as axiom types, synonymity with OWL:Thing, signature difference, etc. Justification corpus based on a large set of ontologies: BioPortal repository, TONES repo, OBO XP. User studies: present set of axioms (justification), a possible consequence and ask user to respond whether the consequence follows from the axioms. Track: answers, time required, eye-track Results: model fairly accurate, but fails when justifications contain superfluous, distracting ISWCparts. 2011 Review
The justificatory structure of the NCBO BioPortal ontologies Samantha Bail et al. Manchester Studies existing ontology corpus (BioPortal) to determine whether entailments with multiple justifications are common in practice. Proposes graph-based framework for describing and analysing relations between justifications in ontologies: bipartite graphs where nodes are axioms or justifications. defines characteristics that can be derived from justification graph (redundancy, activity, axiom-power, self-justifications, isomorphism, etc.) multiple-justifications in 71.4% of ontologies, other measures can help to suggest repairs.
Wheat or chaff Practically feasible interactive ontoloigy revision Nadeschda Nikitina et al. Karlsruhe and Ulm detect incorrect axioms that have been acquired automatically axiom ranking strategies based on logical errors (leading to inconsistency), provenance (whether a human author has validated an axiom). Goal is to minimise validation steps. proposes a ranking algorithm to maximise the impact of an axiom validation proposes a partitioning algorithm to minimise computation load evaluation based on NanOn project (literature search domain)
A novel approach to visualising and navigating ontologies Enrico Motta et al. KMI (Open University), Bologna, isoco Visualisation based on previous work to detect most natural concepts in an ontology Use algorithm to extract the key concepts to provide an improved way to browse through an ontology Evaluation: perform a number predefined tasks using different interfaces (NeOn toolkit with and without KCViz and Protege OWLViz). KCViz resulted in less time required to perform the tasks.
Visualizing ontologies: a case study John Howse et al., Brighton and CSIRO (Australia) Discusses how Concept Diagrams, a variant of Euler diagrams, can be used to visualise ontologies. attractive properties: easy to learn, can be mapped to a large subset of OWL axioms. can be used to visualise entailments and justificatins
Decomposition and modular structure of BioPortal ontologies Chiara del Vescovo, Pavel Klinov et al. Manchester, Arizona, Bremen Introduces notion of Atomic Decomposition Uses atomic decomposition to decompose ontologies in BioPortal corpus and analyses the resulting decomposition Decomposition method very promising for providing fast module extraction for applications since decomposition can be performed beforehand.
Inspecting regularities in ontology design using clustering Eleni Mikroyannidi Introduces a set of primitives for analysing axioms in ontologies such as placeholders for concepts, distance between placeholders, popularity of axioms. Proposes an algorithm for extracting regularities (one or more axioms that occur frequently in an ontology). Shows how the extracted regularities for 4 ontologies can be used to evaluate the ontologies and propose ways to improve the ontology.