=Paper=
{{Paper
|id=None
|storemode=property
|title=A Foundational Ontology to Support Scientific Experiments
|pdfUrl=https://ceur-ws.org/Vol-938/ontobras-most2012_paper12.pdf
|volume=Vol-938
|dblpUrl=https://dblp.org/rec/conf/ontobras/CruzCM12
}}
==A Foundational Ontology to Support Scientific Experiments==
https://ceur-ws.org/Vol-938/ontobras-most2012_paper12.pdf
A Foundational Ontology to Support Scientific Experiments
Sergio Manuel Serra da Cruz1,3, Maria Luiza Machado Campos2, Marta Mattoso1
1
Programa de Engenharia de Sistemas e Computação (PESC/COPPE-UFRJ)
2
Programa de Pós Graduação em Informática (PPGI-UFRJ)
3
Programa de Pós Graduação em Modelagem Matemática e Computacional
(PPGMMC - UFRRJ)
serra@ufrrj.br, mluiza@ufrj.br, marta@cos.ufrj.br
Abstract. Provenance is a term used to describe the history, lineage or origins of a
piece of data. In scientific experiments that are computationally intensive the data
resources are produced in large-scale. Thus, as more scientific data are produced the
importance of tracking and sharing its metadata grows. Therefore, it is desirable to
make it easy to access, share, reuse, integrate and reason. To address these
requirements ontologies can be of use to encode expectations and agreements
concerning provenance metadata reuse and integration. In this paper, we present a
well-founded provenance ontology named Open proVenance Ontology (OvO) which
takes inspiration on three theories: the lifecycle of in silico scientific experiments, the
Open Provenance Model (OPM) and the Unified Foundational Ontology (UFO).
OvO may act as a reference conceptual model that can be used by researchers to
explore the semantics of provenance metadata.
1. Introduction
Currently, researchers are facing a proliferation of a huge volume of scientific data. These
data are often shared and further processed and analyzed among collaborators. There is a
consensus among science data communities that metadata is the foundation for data
discovery, use, and preservation. The importance of managing the provenance metadata of
scientific experiments is becoming vital to researchers as they have to share results and also
consider the long-term usability of data products generated by their investigations.
Provenance captures a derivation history of data products, and is essential to the
long-term preservation, to reuse, and to determine data quality (Freire et al., 2008, Moreau
et al., 2011). Provenance provides transparency in data acquisition and processing,
managing trustworthiness of data sources, allowing those who use the data to determine its
validity and to verify its accuracy. For instance, in scientific domains such as Biology,
Chemistry and Physics, the tendency toward collaborative scientific process is increasingly
evident (Hey, Tansley, Tolle, 2009). Thus, with the proliferation and sharing of such data,
questions such as “where did this data come from?”, “who else is using this data?” and “for
what purpose was it generated?” are becoming increasingly common in the scientific arena.
To ensure that data provided by third-party can be trusted, shared and reused appropriately,
it is imperative that the semantics of provenance is clearly and precisely defined and made
available to users.
Despite of the amount of research papers and surveys about provenance (Buneman
et al., 2001, Oinn et al., 2004, Sahoo et al., 2008, Cruz et al., 2009, Mattoso et al., 2010),
each work describes it according to a different and particular perspective. For instance,
Buneman et al. (2001) and Oinn et al.(2004) describe provenance in terms of common data,
while others describe it in terms of metadata (Cruz et al., 2009 and Mattoso et al., 2010)),
144
i.e., as a secondary piece of information that is complementary in some way to the primary
piece of information to which it refers. Unfortunately, despite of all these research efforts,
scientific data and provenance integration is a problem not completely solved, especially
when it involves semantic issues. With such issues in mind, we are interested in the
semantics of provenance at a novel perspective when compared to other related works
(Salayandia et al., 2006, McGuinness et al., 2007, Sahoo et al., 2008, Zhao, 2010).
The goal of this article is to present and discuss the features of a novel ontology
named Open proVenance Ontology (OvO) which is inspired in three other theories: the
lifecycle of scientific experiments, presented by Mattoso et al. (2010), the Open Provenance
Model (OPM), discussed by Moreau et al. (2011) and the Unified Foundational Ontology
(UFO), proposed by Guizzardi (2005). We advocate that when binding higher levels of
provenance metadata (regarding organization and knowledge about the experiment and its
scientific hypothesis) with fine grained operational provenance (collected during
experiments´ execution) and the explicit specification of the conceptualizations of the in
silico scientific experiments domain; unanswered research questions might be solved by
exploring the semantics of provenance metadata. For instance, one can perform reasoning
about the history of how hypothesis, models, decisions, annotations evolve during recurrent
runs of a workflow that is part of an in silico scientific experiment. Furthermore, one can
explore how an abstract workflow conceived by a researcher is related to a given data
product generated at the laboratory of a research partner.
Differently from related works, in this article we present the conceptual model of a
well-founded provenance ontology about in silico experiments. The purpose of the ontology
is to provide a provenance reference model in order to: (i) support the integration of distinct
kinds of provenance produced during the lifecycle of scientific experiments; (ii) address
semantic interoperability and data/standard integration between experiments; (iii) allow a
novel approach for provenance querying; (iv) convey a knowledge repository about
scientific experiments results; and, finally, (v) represent the provenance of scientific
experiments as a semantic network, exposing it to automated capture and query
mechanisms.
The remainder of this paper is organized as follows: Section 2 provides a brief
background about the role of provenance in the lifecycle of a scientific experiment; Section
3 introduces the ontological engineering approach adopted in this work; Section 4 presents
the OvO ontology; Section 5 discusses related work; and, finally, Section 6 concludes the
paper.
2. Background – The role of provenance on in silico scientific experiments
In silico is an expression used to mean “performed on computer or via computer
simulation”. In silico scientific experiments are characterized by the composition and
execution of several variations of scientific workflows (Mattoso et al., 2009); they are
performed with the aid of computer systems and provide researchers a number of
advantages, such as: higher precision and better quality of experimental data; better support
for data-intensive research and access to vast sets of experimental data generated by
scientific communities; more accurate simulations through scientific workflows and higher
productivity. However, in silico experiments nowadays suffer from an increasing
complexity on setting up, maintaining and making changes to the simulation systems and
also the shortcomings of managing large data sets of experimental data and provenance
metadata.
Like traditional scientific experiments, in silico scientific experiments, independent
of their domain, follow some common directions (Jarrad, 2001) such as: (i) they need to be
145
re-executed, enabling other researchers to conduct similar experiments to confirm (or
refute) the results; (ii) the results need to be well documented to be used as a baseline for
further experiments, conducted by different researchers in different laboratories; (iii) they
must follow a protocol or a methodology and be executed under controlled conditions.
However, in silico experiments are often specified and materialized as scientific workflows.
Scientific workflows are defined in two levels. In a higher level of abstraction, an
abstract workflow is described as conceptual model without binding to specific
computational resources. In the lower level, a concrete workflow binds programs and data
allowing the structured composition of a sequence of operations aiming its execution to
achieve a desired scientific result. To be effectively managed the scientific workflows
require a specific set of cardinal facilities, such as experiment specification techniques,
workflow derivation heuristics, provenance gathering mechanisms and high performance
computing environments ranging from private clouds, commercial clouds such as Amazon,
GoGrid, Rackspace to supercomputing centers (Stevens et al., 2007, Mattoso et al., 2010).
Manual analysis of the results data set of in silico experiments is commonly
unfeasible. This involves, for instance, checking input and output data sets of each activity
of the workflow, verifying if computations failed on remote computational resources, and
checking all activities that contributed to the creation of a particular data set. Many of these
activities can be automatically executed by querying provenance metadata.
The treatment of provenance as a first-class data artifact was primarily introduced at
OPM by Moreau et al. (2011). The OPM is a way of recording information about artifacts,
agents and processes as they occur which includes constructs for representing causal and
dependency relationships between sub-processes and the data items or other artifacts that
they use or produce. The OPM is a provenance metamodel which is gaining popularity and
for which implementations are becoming available in OWL, RDF and Java. Despite of its
increasing popularity OPM has some issues, for example, it neither considers all kinds of
provenance of in silico experiments nor expresses an unambiguous semantic model.
There are two types of provenance. Prospective provenance describes how these
scientific workflows were composed and Retrospective provenance describes how they
were executed and also the relationships between the input and output data sets (Freire et
al., 2008). Another important characteristic of provenance is related with its granularity
(Cruz et al., 2009), also referred to as provenance level. It is generally classified as coarse
or fine-grained. The desirable level of provenance granularity depends on application
domain requirements, and the cost of collection, storage and processing. The finer the
granularity of the provenance record, the higher the information entropy and associated
cost.
According to Mattoso et al. (2010), in silico scientific experiments can be described
as being part of a lifecycle, which consists of three stages: Composition, Execution and
Analysis. Each stage has an independent cycle, taking place at distinct moments of time and
handling different kinds of provenance metadata. At the Composition stage, researchers
either elicit the requirements to build a new abstract workflow as software or retrieve old
ones to reuse for new purposes. They start the composition process building the raw version
of the workflow by choosing programs incrementally, backward or forward, along the stage.
Then, they refine the successive versions towards a concrete workflow.
At the Execution stage researchers make their essays by executing instances of a
concrete workflow according to their own needs using real parameters and data sets in a
production environment. Researchers can also monitor (local or distributed) experiments
and register retrospective provenance metadata that can be further used in debugging or
146
optimization activities, e.g., researchers may optimize concrete workflows, this can be
obtained by usage of parallelism and distributed computation. They are also able to initiate
an analytical process undertaken over outcomes of collections of experiment runs,
Finally, at the Analysis stage researchers can investigate the data products generated
by the experiment and then publish results or share not only its outcomes, but also the
whole workflow or its parts to other domain experts. This stage may be further decomposed
to support the analysis of results. The researchers may face two different situations when
analyzing the results: (i) with the generated data, they may conclude that results are likely to
be correct, but decide to continue with the experiment, varying parameters and ingesting
others data sets, to prove or refute the hypothesis; (ii) when analyzing the data products,
researchers discover that their hypothesis is refuted, but faces a new scenario that may lead
to a new discover, thus raising a new hypothesis.
In this article we advocate that in silico scientific experiments can be fully described
as hierarchical trails of provenance metadata with different kinds of granularity collected
during its lifecycle. First, at a higher and abstract level, we have the prospective
provenance; it cannot be gathered by the existing Scientific Workflow Management
Systems (SWfMS); at a lower level we have fine grained provenance collected during the
execution and analysis of the experiments.
3. The Ontological Engineering Approach
The first version of the Open proVenance Ontology (OvO) was initially developed by Cruz
(2011). The author adopted a combination of two methodologies found in the literature.
Firstly, we have employed the ontology engineering process which includes the phases of
conceptual modeling, design and codification (Gomez-Perez et al., 2004 and Guizzardi,
2007). The conceptual modeling of ontologies should strive for expressivity, clarity and
truthfulness in representing the subject domain at hand. Due to space restrictions, the
methodology and the rationale behind of the OvO´s conceptual modeling and design will
not be fully discussed in this paper. A detailed description can be found at Cruz (2011).
As second methodology, we have used the ontologically well-founded UML
modeling profile proposed by Guizzardi and Wagner (2005). This profile comprises a
number of stereotyped classes and relations implanting a metamodel that reflect the
structure and axiomatization of a foundational and domain independent ontology named
Unified Foundation Ontology (UFO). The UFO was stratified in three ontological layers
(fragments), namely UFO-A, UFO-B and UFO-C. A complete description of UFO falls
completely outside the scope of this paper. However, we give an overview of the fragments
of this ontology which were used in the instances of the modeling profile employed in this
article.
UFO-A is the core of UFO, it defines concepts related to endurants. An endurant is
an entity that is present as a whole at any given point in time, i.e., it does not have temporal
parts and persists in time while keeping its unique identity. The endurants (universals and
particulars) can be summarized as follows: Kinds are rigid, independent sortals that supply a
principle of identity for their instances; Phases are independent anti-rigid sortals; Roles are
anti-rigid and relationally dependent sortals, RoleMixins are non-sortals that can be
partitioned into disjoint subtypes which are sortals. Examples of endurants are a laboratory,
an organization, a researcher, an experiment and its phases. UFO-A also uses relations.
Relations are entities that connect other entities together, they are divided as follows:
Formal relations hold between two or more entities directly, without any further intervening
individual. Examples of formal relations include “a laboratory has projects”. Contrarily,
material relations have a material structure of their own and have their relata mediated by
147
individuals called relators. For instance, a workflow_run is a relator that connects a concrete
workflow and an executor.
UFO-B builds upon UFO-A and defines concepts related to perdurants. A perdurant
is an entity composed of temporal parts, i.e., its existence extends in time accumulating
temporal parts. Examples of perdurants are the codification of a concrete workflow and the
composition of an abstract workflow. It is not always the case that whenever a perdurant is
present all of its temporal parts are also present. With a perdurant, if we freeze time we can
only see a limited number of parts of the perdurant and not the whole. For instance, in a
“coding session" of an concrete workflow, if we take a snapshot at the point in time when
the programmer is having its code validated in order to execute a concrete workflow, we
cannot determine that this task is part of the “execution" of the scientific experiment.
UFO-C defines social and intention-related concepts (both endurants and
perdurants) and is built on top of UFO-A and UFO-B. One of the main distinctions made in
UFO-C is between Agents and Objects. An Agent is a substantial that creates actions,
perceives events and to which we can ascribe mental states (intentional moments). Agents
can be physical (a football player) or social (a stadium). An Object is a substantial unable to
perceive events or to have intentional moments. Objects can also be further categorized into
physical (e.g. a ball) and social objects (e.g., money).
4. Open Provenance Ontology1
In this section we discuss the key concepts of Open Provenance Ontology (OvO). OvO´s
concepts are depicted as UML class diagrams because of the widespread understanding of
UML classes and relations and their suitability for our purposes of illustrating how the
presented concepts are organized and how they are related to each other. OvO was
developed as a set of three sub-ontologies: (i) in silico scientific experiment sub-ontology,
(ii) experiment composition sub-ontology, (iii) experiment execution sub-ontology. The sub-
ontologies complement each other; they are connected by relations between their concepts
as well as by formal axioms.
The UFO stereotypes of the modeling profile used are signed between the sign
<< >> and the names of the concepts of the ontology (classes) are typed in italics. The
classes are colored to facilitate understanding. The concepts colored in yellow map
prospective provenance and in green retrospective provenance. Due to space restrictions,
solely the key concepts are discussed, the complete description of all concepts of OvO can
be found at (Cruz et al., 2009 and Cruz, 2011).
4.1. The in silico scientific experiment sub-ontology
This sub-ontology captures the structure of in silico scientific experiments at a high
abstraction level (Figure 1).
An in silico experiment is a research that has the purpose of discovering something
unknown by adopting a computational model and by the evaluation of a hypothesis that
involves the design and implementation of an in silico experiment. Experiment and Project
are labeled as <> that is a rigid sortal universal that can be identified in all possible
worlds. We consider in silico experiments as a composition of three essential parts: the
model, the hypothesis and project (all are represented by the essential=true tagged value in
the part-whole relation notion); this relationship indicates that an experiment is made of
parts and inseparable relationally dependent. Hypothesis and Model are also rigid sortal
1
Only key concepts and relationships were illustrated and discussed.
148
universals, represented as disjoint concepts tagged as <>. The models do not exist
outside the context of the experiment as a whole. They are inseparable parts of the scientific
experiment they comprise (represented with the mark inseparable=true). The Model defines
the limits of a scientific investigation, recording the circumstances surrounding an
experimental study. The Hypothesis is a proposition that is stated in an attempt to verify the
validity of a provisional response.
Project refers to the research project in which the in silico experiments are
conducted. A project defines the location of the research and the involvement of several
types of researchers. Project is represented as an hierarchy. The Laboratory and
Organization concepts are rigid sortals and are related by aggregation and composition with
the underlying concepts. The Laboratory uniquely identifies the points in geographical
space where a research project is designed and where the in silico experiments are
conceived.
Figure 1. Fragment of the in silico scientific experiment sub-ontology
The Organization is a rigid sortal universal which plays an important role in the
sub-ontology. For example, COPPE/UFRJ is an organization that has several laboratories
(e.g., databases, software engineering, among others) and they have material and human
resources that can be allocated to conduct a research project. Laboratory and Organization
are independent, but complementary, one can change the members of an organization
(university, laboratory) without losing its identity principle. The Laboratory also has
another important feature, it stresses the relationship with Person as an aggregation (tagged
as a rigid sortal). Here, we assume that being a researcher is an extrinsic property of a
person, i.e, there are worlds in which a person is not a researcher and however, he still
remains a person. Person is regarded to any human being, however, only those who plays
any role in conducting a scientific experiment are mapped as Researcher. To represent the
possible roles played by people whose main attribute is to be a researcher, we adopt the
stereotype <> of the UFO-A, representing them as anti-rigid and relationally
dependent sortals. Each instance of researcher must be an instance of person who inherits its
identity principle. The concepts Executor, Programmer, Architect and Coordinator are sub-
types of Researcher.
Returning to Experiment, it is tagged as rigid sortal and its graphical representation
is the same as a generalization of the UML metamodel. However, we can notice different
semantics. For instance, in UML, the classes to a generalization are necessarily disjoint and
by default do not form a partition. Taking into account the lifecycle of in silico experiments
(as described in Section 2) and the theory behind UFO-A. An experiment may be
represented as a disjoint set (label {disjoint, complete}) of stages. It can be decomposed by
149
two different concepts: Composition_phase and Execution_phase. Both are tagged as
<> which means that each stage of the lifecycle is an independent anti-rigid sortal
that happens in different points in time and represents a condition that depends solely on its
intrinsic properties.
4.2. The experiment composition sub-ontology
This sub-ontology represents prospective provenance associated to the composition stage of
an in silico experiment (Figure 2). The concepts related to prospective provenance are
yellow colored while the ones related to retrospective provenance are green.
The composition stage of in silico experiments comprises all tasks of specification
and modeling of abstract and concrete workflows and their activities. During the modeling
task, we capture the knowledge related to the materialization of the experiment and the
design of the scientific workflows. At the highest level of the metamodel there is the
concept Composition_Phase. Such concept is tagged as anti-rigid sortal. By this kind of
representation we ensure the unique identification of each cycle of composition in the life
cycle of the experiment executed in a given organization.
The association between Composition_Phase and Workflow has an important
semantic explanation; it allows the specialization of the two types of workflows found on in
silico experiments: abstract and concrete workflows. Workflow is tagged as <>, which means that such kind of event is composed of other events by means of
event composition operators.
UFO-B events are possible transformations from a portion of reality to another, i.e.,
they may change the reality by changing the state of affairs from one (pre-state) situation to
a (poststate) situation. These are complex entities that are constituted by possibly many
endurants. Situations are taken to be synonymous to what is named state of affairs in the
literature, i.e., a portion of reality which can be comprehended as a whole. According to our
metamodel, the concepts Workflow_Code and Workflow_Description are tagged as
<>, such approach allow us to uniquely identify the versions from one prior-version
of a workflow (pre-state) to a novel-version (poststate) of a workflow.
Complex events are aggregations of at least two perdurants (either atomic or
complex events). Perdurants are ontologically dependent entities, i.e., perdurants
existentially depend on their participants in order to exist. The Concrete_workflow only
exists with the participation of the Workflow_code, the Concrete_activity and the
Atomic_Concrete_Activity (the substantial in the model). Similar arguments can be used for
Abstract_workflow.
Concrete_activity and Abstract_activity are also complex events. For instance,
Concrete_activity is modeled using the weak supplementation pattern (Guizzardi, 2005).
The pattern represents a parthood where the hollow diamond is connected to the whole
(Complex_Concrete_Activity) and the aggregation is supposed to be a irreflexive and anti-
symmetric relation.
The specialization of the concept Researcher (discussed in section 4.1) in distinct
roles, i.e. relationally dependent sortals, is crucial to expose the different duties of the
members of a research team during the lifecycle of the experiment. For example, the
architect of the experiment is a highly trained domain researcher who is in charge of
planning, design and oversight/supervision of the experiment, he may not worry about the
software packages versions involved, technical and operational details needed to build
concrete workflows. His duties are related to the composition of the sequence of activities
of an abstract workflow, while the programmer is related to technicalities and the
150
production of executable codes. He is the person who writes concrete workflow as
computer software. The roles represented by Architect and Programmer are tagged as
<> having two explicit relationships. The first refers to actions that happen in time,
the actions Coding_Session and Design_Session are tagged as <>, the second
relationship involves the rigid sortal SWfMS.
Figure 2. Fragment of the experiment composition sub-ontology
The semantics of Coding_Session and Design_Session exposes different events that
happen in time (the act of coding a concrete workflow or designing an abstract workflow
respectively). An Action is a UFO-C perdurant that is an individual instance of an Action
Universal, with the purpose of satisfying the propositional content of an intention. Actions
are intentional events, i.e., events which instantiate a plan with the specific purpose of
satisfying some internal commitment of entities capable of bearing intentional moments.
C_Situation and A-Situation are tagged as UFO-C <>.
4.3. The execution experiment sub-ontology
This sub-ontology represents retrospective provenance associated to the execution stage of
an in silico experiment (Figure 3). The concepts related to retrospective provenance are
green colored.
At the highest level there is Execution_Phase tagged as a universal sortal
<>; such representation ensures the unique identification to each execution of a
concrete workflow during the lifecycle of the experiment. During the execution stage, the
researcher can execute different instances of a concrete workflow. Therefore, the specific
workflows of an experiment are associated by composition to the execution stage.
A concrete workflow comprises of one or more concrete activities. The
Concrete_Workflow2 represents the association, by composition, to the concept
Concrete_Activity. A concrete activity can be understood as a codification of an executable
scientific application individually (Program) which shall be governed by the principle of
unique identification.
2
In order to simplify the diagram of the experiment execution sub-ontology (Figure 3), the weak
supplementation pattern and the C_Situation on Concrete_Activity were omitted.
151
The Artifact is a piece of data represented by a rigid sortal. For instance, a
researcher Diogo uses a resource “FastaFile_TripCruzi_20122010.txt” which is owned by
researcher Alberto. All artifacts, individually, must have their own principle of identity and
also additional essential properties, such as name, type and date of creation, among others.
The artifacts (pieces of data) handled by a specific concrete workflow, can be specialized
as: Input_Data, Output_Data and Parameter which are shown as <>.
It is worth mentioning that associations between Artifact and Concrete_Activity
have a particular meaning. The association usedBy, generatedBy, and triggeredBy,
derivedBy (shown in blue color in Figure 3) are inherited from the OPM metamodel
(Moreau et al., 2011). The association describes the artifacts which were consumed by a
given process (concrete activity), while the second describes those that were produced
during the processing of a given activity. Finally, the third represents the sequence of the
specific activities of a particular concrete workflow. Strictly speaking, the associations
originally defined by OPM have semantic meanings that could be better explained. For
example, usedBy was plotted as a simple association. However, if we consider a
semantically rigorous model, the association between Concrete_Activity and Artifact can be
represented through a third-class R´ type material relator to explain all the details of the
semantics of this relationship.
The execution of the concrete activities of a workflow is performed in a
computational environment, so it is necessary to expose this association. In Figure 3, we use
a combination of Concrete_Activity and Environment. However, an execution environment
is not a monolithic entity, by the contrary; it is represented by an aggregation of the rigid
sortals Program and Hardware. We have described the succession of resources involved in
the execution of a specific workflow. In this case, Resource is tagged as a non-sortal
<> type of UFO-A. That is, an abstract class that allows for specialization of
other classes and subsequent identification of these resources, as Hardware_Resource,
Software_Resource, and Researcher. We decided to use the term Researcher rather than,
Human Resource, since the second term represents the entire set of employees of an
organization, while the first represents the subset of employees who are trained and
qualified to conduct scientific research. Hardware_Resource and Software_Resource are
tagged as <>.
Finally, it is important to stress the material relations depicted in Figure 3. The
explicit association between Executor and Concrete_Workflow is represented by an n-ary
relationship. In this case, such relationship is mapped as Workflow_Run, representing
sessions of work performed by the executor during the execution of a concrete model of a
scientific workflow. In light of the UFO-A, universal relators are represented by the
stereotype <> and material relations are represented through associations
stereotyped as <>. The dotted line relationship between a material relationship
and the universal relator indicates that the relationship is derived from a material relator.
The presence of the graphic symbol (●) on the relator (at the end of the dotted line) is used
to distinguish the graphical representation of a class of a simple UML association class. The
relationship between the relator Workflow_Run, Concrete_Workflow and the Executor
tagged as an anti-rigid <> is an example that represents the above mentioned
material relationship. Textually, we have the following: Workflow_Run (X, Y) is true if and
only if there is an Executor X and a Concrete_Workflow Y.
152
Figure 3. Fragment of the experiment execution sub-ontology
5. Related Work
There are few research initiatives committed with ontology-based models and
formalizations of the in silico scientific experiments. A number of provenance ontologies
have emerged from scratch or through conversion of existing provenance vocabularies. The
main features with these provenance ontologies are: (i) they are focused on the provenance
of digital objects (especially data) not the experiment; (ii) they leverage only the Web as the
underlying infrastructure for data generation and data access; (iii) they encode
computational semantics for provenance vocabularies using declarative representations; and
finally (iv) they do not take into account a formal method like the one based on
foundational ontology principles to validate its development.
The Open Provenance Model Ontology (Zhao, 2010) focuses solely on retrospective
provenance of concrete scientific workflows. It defines a small set of core concepts for
general entities (artifacts, agents, and processes) and relations in workflows, e.g., “artifact
wasGeneratedBy process” and “process wasControledBy agent”. With a small number of
concepts to represent such a complex domain, it results in concepts being semantically
overloaded. It was developed by a community of workflow researchers and has multiple
serialization profiles, including XML Schema and OWL. Currently, the OWL profile is still
evolving to adapt the OPM specification to be a W3C standard.
The Proof Markup Language (PML) (McGuinness et al., 2007) focuses on
information manipulation processes such as logical reasoning, information extraction, and
more recently, machine learning. It is modularized into several loosely coupled modules to:
(i) annotate provenance metadata for sources of knowledge, (ii) encode information
manipulation processes and data dependencies for deriving the conclusions or executing
workflows and (iii) annotate trustworthiness assertions about knowledge and sources. PML
uses a proof theoretic foundation for its model, its specification is synchronized with its
own OWL ontology.
153
The Workflow Driven Ontology (Salayandia et al., 2006) uses PML to represent
abstract workflows, which is a plan for a workflow but not the execution of a workflow. It
uses the Method concept to represent the class of actions to be executed and uses the Data
concept to represent the class of data to be operated on by an action in the workflow. The
Provenance Vocabulary (Hartig, Zhao, 2010) focuses on information manipulation. It
consists of three modules: the core module defines basic concepts for representing data
creation and data access processes, and the other two modules extend the core module by (i)
adding classifications specific to Web information transfer and (ii) supporting
authentication of information. It should be noted that this ontology uses OWL 2 language
features. Provenir (Sahoo, Seth, 2009) is an ontology based on information manipulation. It
reuses and redefines some provenance relations from the OBO Relation Ontology (Smith et
al., 2007), which defines generic binary relations without domain/range specifications.
Provenir, like the previously mentioned works, (re)defines some of the universal
OPM classes (process, artifacts and agent) to its own purposes. These ontologies do not
consider the role and the granularity of the different kinds of provenance metadata
generated during the complete lifecycle of in silico experiments. However, as discussed,
they tend to be too much driven by both requirements of specific applications and less
committed to maximizing expressivity, clarity and truthfulness with regard to the domain of
generic in silico scientific experiments. Furthermore, as a rule, such initiatives are not
committed to the use of top-level ontologies such as UFO and, consequently, do not take
advantage of neither its ontological foundation nor its subjacent ontological engineering
approach to create clear conceptual models of high expressivity.
6. Conclusion
This article has presented a novel provenance ontology name Open Provenance Ontology.
We advocate that the theory behind the OvO provides: (i) a knowledge repository about the
different kinds of provenance of in silico scientific experiments and (ii) a reference
conceptual model that may be used for promoting interoperability between distinct
provenance systems.
Our contributions cover a gap in literature with regard to ontological approaches for
modeling provenance metadata of scientific experiments. However, there is a need for
future research on suitable languages to model phenomena such as the dynamics of the
composition and execution of workflows in distributed environments. At this time, OvO is
part of a distributed provenance gathering system named Matriohska, its modeling is being
reviewed and part of it, implemented as OWL, and is under evaluation with in silico
bioinformatics experiments at Fiocruz. Further details can be found at Cruz (2001).
Additional future work includes: (i) the design and implementation of the OvO ontology in
one or more codification languages (e.g., F-Logic) as a proof of concept and (ii) the
extension of the ontology for covering as far as possible novel situations.
References
Buneman, P., Khanna, S., W-C Tan, (2001) “Why and Where: A Characterization of Data
Provenance”. In LNCS v. 1973 n. 2001 p. 316-330.
Cruz, S. M. S., Campos, M. L. M., Mattoso, M. (2009) “Towards a Taxonomy of Provenance in
Scientific Workflow Management Systems”. In: SERVICES '09, p. 259-266.
Cruz, S.M.S., (2011) “Uma Estratégia De Apoio À Gerência De Dados De Proveniência Em
Experimentos Científicos” Tese de Doutorado, PESC/COPPE-UFRJ , 234p.
Freire, J., Koop, D., Santos, E., Silva, C. T. (2008) "Provenance for Computational Tasks: A
Survey," Computing in Science and Engineering, v. 10, n. 3, p. 11-21.
154
Gomez-Perez, A., Corcho, O., Fernandez-Lopez, M. (2004) “Ontological Engineering : with
examples from the areas of Knowledge Management, e-Commerce and the Semantic Web. 1
ed. Springer, 2004.
Guizzardi, G. Wagner, G. (2005) “Some applications of a unified foundational ontology in
business modeling”. Business Systems Analysis with Ontologies, chapter 13, p. 345–367.
Guizzardi, G. (2005)"Ontological Foundations for Structural Conceptual Models" CTIT PhD.-
thesis series, No. 05-74, 441p.
Guizzardi, G. (2007). “On Ontology, ontologies, Conceptualizations, Modeling Languages, and
(Meta)Models”. In Frontiers in Artifical Intelligence and Applications, Databases and
Information Systems IV, IOS Press, Amsterdam.
Hartig, O., Zhao, J., Provenance Vocabulary Core Ontology Specification. (2010). Available at:
.
Hey, T., Tansley, S., Tolle, K., (2009) “The Fourth Paradigm: Data-Intensive Scientific
Discovery”. 1. ed. Redmond, Microsoft Research.
Jarrard, R. D. Scientific Methods, an online book. 1. ed. Dept. of Geology and Geophysics,
University of Utah, 2001.
Mcguinness, D., Ding, L., Silva, P. P. (2007) “PML 2: A Modular Explanation Interlingua”. In:
Proc of the 2007 Workshop on Explanation-aware Computing, pp. 49-55.
Mattoso, M., et al.. (2010) "Towards Supporting the Life Cycle of Large Scale Scientific
Experiments". International Journal of Business Process Integration and Management, v. 5,
p. 79-92.
Moreau, L., et al.., (2011) "The Open Provenance Model core specification (v1.1)". Future
Generation Computer Systems, v. 27, n. 6, pp. 743-756.
Oinn, T. M., Addis, M., Ferris, J., Marvin, D., Senger, M., Greenwood, R.M., Carver, T.,
Glover, K., Pocock, M. R., Wipat, A. P. Li. (2004) “Taverna: a tool for the composition and
enactment of bioinformatics workflows”. Bioinformatics, 20(17):3045– 3054.
Sahoo, S. S., Sheth, A., Corey, H. (2008) “Semantic Provenance for eScience: Managing the
Deluge of Scientific Data” Internet Computing, IEEE, v. 12 n. 4, p. 46-54.
Sahoo, S. S., Shet, A., (2009) "Provenir ontology: Towards a Framework for eScience
Provenance Management". In: Microsoft eScience Workshop.
Salayandia, L., da Silva, P.P. Gates, A. Q., Salcedo, G. F. (2006) “Workflow-Driven
Ontologies: An Earth Sciences Case Study”. In E-SCIENCE '06 Proc. of the 2nd IEEE
International Conference on e-Science and Grid Computing.
Stevens, R., Zhao, J., Goble, C, (2007) “Using provenance to manage knowledge of in silico
experiments”, Brief Bioinformatics, v.8. n. 3, p. 183-194.
Smith, B., et al. (2007) The OBO Foundry: coordinated evolution of ontologies to support
biomedical data integration. Nature Biotechnology 25, 1251–1255.
Zhao, J. Open Provenance Model Vocabulary Specification. (2010). Available at .
155