|
The main goal of the team is to gain users' confidence
in workflow execution results by enhancing results with
provenance information, trust recommendations, and levels
of uncertainty. Provenance, trust, and uncertainty about
the results of cyber-infrastructure-based applications,
e.g., scientific workflows, are essential if scientists
are to believe and, thus accept, these results. This subproject
will address the complex problem of using provenance as
the key enabler for integrating trust management and uncertainty
management in distributed environments like the grid. The
subproject will support research leading to a uniform way
of representing uncertainty and trust. Uncertainty models
embedded in provenance will be comprehensive enough to support
the computation of several dimensions of uncertainty including
error, accuracy, and reliability. Trust models embedded
in provenance will be rich enough to support the computation
of trust recommendations that can describe several aspects
of trust including distrust, partial trust, ignorance, and
inconsistencies.
Team
- Paulo Pinheiro da Silva (Lead, CS)
- Vladik Kreinovich (CS)
- Ann Gates (CS)
- Leonardo Salayandia (Research staff – Center Team)
- Nicholas Del Rio (Ph.D. student)
- Aida Gandara (Ph.D. student)
- Alejandro Castaneda (M.Sc. student)
Summary of Year 1 Activities
Provenance. Activities centered on the
development of a lightweight Proof Markup Language which
will encode basic justifications and can be used by a restricted
set of Inference Web tools for retrieval and browsing of
provenance information.
Trust. The focus was the development of
TrustMap, a strategy for propagating and aggregating trust
about a scientific artifact. TrustMap can facilitate the
acceptance of an artifact as a quality product by its users.
CI-Learner, another approach developed to extract a social
trust network from information available on the web, considers
co-authorship as an explicit evidence of trust between agents
of the network, and it limits its scope to trust relations
that can be extracted from one scientific portal. An IRIS
use case scenario was developed and used to extract trust
network information for the Earth Science community. Activities
demonstrated the potential of providing trust recommendations
through trust maps and the use of data to construct a social
trust network. A user study assessed completeness and usefulness
of ranking as perceived by members of the network.
Workflow. Abstract workflows in the form
of workflow-driven ontologies (WDOs) were used to document
and design workflow artifacts. An abstract workflow can
be manipulated by domain experts because abstraction hides
implementation details irrelevant to the experts. As a result,
domain experts can better understand and independently enhance
the workflow.
The WDO approach has been formalized and WDO-specific tools
have been developed to replace generic tools such as OWL
editors and spreadsheets. With the introduction of an upper-level
class hierarchy of workflow-related classes, WDOs facilitate
the process of creating ontologies to be used on scientific
domains. With the introduction of a well-defined set of
relationships between classes, WDOs are used to guide scientists
through the process of relating classes in a way that can
later be polished into useful workflow specifications. With
the introduction of indirect relationships, WDOs enable
knowledge capture for underspecified processes that can
be refined later.
Uncertainty. Activities focused on developing
new and faster algorithms for processing uncertainty related
to cyberinfrastructure where different data points come
with different, e.g., probabilistic, interval, and fuzzy,
information about their uncertainty.
Summary of Year 2 Activities
The team started implementing systematic semantic enhancements
to geoscience and environmental science processes to improve
the problem-solving capabilities of these processes. The
enhancements are part of the subproject goal of working
collaboratively with geoscientists and environmental scientists
to accomplish the following objectives:
- To understand how scientists tackle scientific problems,
and in particular, how they develop, execute and improve
complex processes to solve those problems;
- To expose scientists to semantic and cyber-infrastructure
technologies that have the potential of overcoming known
limitations of the current scientific processes;
- To indentify key enhancements that can better support
solving the problems identified in #1;
- To develop tools that can be used to create and
maintain artifacts in support of enhancements identified
in #2; and
- Using the proposed enhancements in #3 and the tools
in #4, to collaboratively work with scientists to enhance
their complex processes.
These objectives led to documenting selected scientific
processes. A goal is to establish a new Cyber-ShARE-way
of enhancing these scientific processes with semantic information,
the CI-Miner approach, and to develop human resources in
other scientific fields capable of replicating and disseminating
the use of semantic technology. The following processes
were documented with semantic enhancement:
- Hole’s Code – Earth science/seismology
- Reflectance Data Gathering – Environmental science
- 2 ½ D crustal structure of the Earth – Earth science
- National Center for Atmospheric Research (NCAR)
CHIP Quick Look Process – Solar physics
More efficient algorithms were developed for processing
uncertainty, especially uncertainty related to cyber-infrastructure,
where different data points have different information about
their uncertainty.
|
