HR: 1340h
AN: IN43A-0326 [Abstracts]
TI: DataFinder: Using Ontologies and Reasoning to Enhance Metadata Search
AU: * Russ, T A
EM: tar@isi.edu
AF: University of Southern California, Information Sciences Institute
4676 Admiralty Way, Marina del Rey, CA 90292-6695
United States
AU: Chalupsky, H
EM: hans@isi.edu
AF: University of Southern California, Information Sciences Institute
4676 Admiralty Way, Marina del Rey, CA 90292-6695
United States
AB:
The Southern California Earthquake Center's Community Modeling
Environment uses computer codes for simulation and hazard analysis
computations. The process of running workflows using several
computational models produces numerous intermediate and final data
files. These files have descriptive metadata stored as pairs of
attribute names and values. Depending on which software was used to
prepare the files, different attribute names and different
organizational schemes are used for the metadata. Previous search
tools for this metadata repository rely on the user knowing the
structure and names of the metadata attributes in order to find stored
information. Matters are made even harder because sometimes the type
of information in a data file must be inferred. For example, seismic
hazard maps are described simply as ``JPEGFile'', with the domain
content of the file inferable only by looking at the workflow that
produced the file. This greatly limits the ability to actually find
data of interest.
DataFinder uses ontologies to provide a semantic overlay for the
metadata attributes that are used to index data files. A domain
ontology is combined with a metadata attribute ontology to link
geophysical and seismic hazard domain concepts with the metadata
attributes that describe the computational products. DataFinder uses
a domain ontology and additional rules expressed in first-order logic
to provide this semantic enhancement. The domain and metadata
attribute ontology is represented in the PowerLoom representation
language. DataFinder is implemented using a hybrid reasoning approach
based on combining the strengths of the PowerLoom logical reasoning
engine with the database technology underlying the metadata
repository to provide scalability.
The PowerLoom reasoning engine allows to add semantic enhancements by
overlaying the raw metadata with a hierarchy of concepts, providing
more abstract views of the data collection. For example, a velocity
mesh is one of the intermediate data products used in seismic wave
propagation simulations. It is the output of a 3-dimensional wave
propagation velocity model of the subsurface geology. This mesh can
be computed using any one of several models which have different
characteristics. Each velocity mesh has metadata that records the
specific model used. But the models also fall into general classes
such as 1-dimensional and 3-dimensional models. Such classification
of models can be used to allow retrieval of velocity meshes created by
any 3-dimensional model, without the user being required to specify
all of the particular model names. This provides an abstraction over
the attributes and values stored with the dataset. Other mappings are
used to present a uniform interface to metadata information that is
stored using different attribute names or even different
organizational schemes. Some data products produced as the output of
a computational workflow retain all of the metadata information,
whereas others rely on following a chain of references to the
intermediate products. By using an ontology as an abstraction layer,
DataFinder thus allows users to locate end data products using
domain-level descriptions instead of program-specific and varied
metadata attributes.
DE: 0525 Data management
DE: 7290 Computational seismology
SC: Earth and Space Science Informatics [IN]
MN: Fall Meeting 2005