eScience for Sea Science: A Semantic Scientific Knowledge Infrastructure for Marine Scientists Dr Kristin Stock Allworlds Geothinking Centre for Geospatial Science, University of Nottingham EDINA, University of Edinburgh
Introduction Scientists traditionally search for resources using keyword searches. COMPASS (the COastal and Marine Perception Application for Scientific Scholarship) is a knowledge infrastructure and interface that provides enhanced discovery. Employs semantic information and scientific knowledge.
Goals Enhanced discovery. Advanced visualisation and inference. Semantic architecture: Open standards Interoperability Services Oriented Architecture (aka SDI). Formal ontologies Informal user tagging. Different types of resources: Publications Data Web services
Outline Architecture User Interface Evaluation
The COMPASS Project Funded by UK Joint Information Services Committee Exploring the use of semantic approaches to discovery and access of resources Partners: EDINA Allworlds Geothinking University of Muenster DERI, National University of Ireland, Galway Geosciences, University of Edinburgh Finished June 2009
Related Work (1) Keyword search: Google Scholar, Online collaboratories. Do not allow for search based on scientific methods, theories, etc. Do not allow for geographic or temporal search. Do not allow all of these to be combined for discovery and visualisation.
Related Work (2) Foundation work: models of scientific knowledge and discovery (Hars) Turned into an ontology (SKIo) by Brodaric et al. Science Commons focuses on security, copyright, licensing. GEON Geosciences Network – develops a cyberinfrastructure. People can publish and share resources (data, tools, web services, knowledge). Supports collaborative research. Social networking tools (e.g. ResearchGate) includes semantic search.
But... None of these has provided advanced discovery based on science, with ontologies and interoperability. Hence COMPASS...
Software Architecture Geospatial registry standards: CSW Registry interface. No semantic content usually (ebRIM, metadata). Created an OWL application profile for CSW Keeps the ontologies in their native form: Not done like this previously. Information model = RDF and OWL information models. Provides standard interface to the ontologies. Existing reasoning tools can still be used.
What does this architecture achieve? Interoperability through open standards (registry interface) Between different CSW registries With digital library standards. Semantic Stores OWL ontologies.
Information Architecture All content is in ontologies. Avoids need for two different structures. Several different ontologies used.
Evaluation: User Experience Qualitative evaluation with 12 users. Used the interface and then: completed the MS Desirability Toolkit semi-structured interview Discovery by user tagging (the old way) had the most positive response. Discovery by domain ontology positive (liked graphical display, connections). Discovery by scientific knowledge ontology not well understood.
Evaluation: Data Population Need for ontologies a shortcoming Domain ontology creation: a lot of work! Application ontologies also a lot of work Perhaps some parts can be automated But need to understand the science Best done by the scientist who created the resource.
Evaluation: Architecture Architecture worked well. Some effort to make ontologies fit registry interface. OWL-S very cumbersome.
Conclusions Showed how scientific knowledge can be used for discovery. Generally supported but some improvement needed. More work needed to help with data population and improve user experience. Future potential: Visualisation of scientific patterns Inference to find trends http://compass.edina.ac.uk/beta/csw-broker/