=Paper=
{{Paper
|id=Vol-2969/paper64-OntoCom
|storemode=property
|title=Towards an Ontology Network for the Reproducibility of Scientific Studies
|pdfUrl=https://ceur-ws.org/Vol-2969/paper64-OntoCom.pdf
|volume=Vol-2969
|authors=Sheeba Samuel,Alsayed Algergawy,Birgitta König-Ries
|dblpUrl=https://dblp.org/rec/conf/jowo/SamuelAK21
}}
==Towards an Ontology Network for the Reproducibility of Scientific Studies==
https://ceur-ws.org/Vol-2969/paper64-OntoCom.pdf
Towards an Ontology Network for the
Reproducibility of Scientific Studies
Sheeba Samuel1,2 , Alsayed Algergawy1 and Birgitta König-Ries1,2
1
Heinz Nixdorf Chair for Distributed Information Systems, Friedrich Schiller University Jena, Germany
2
Michael Stifel Center Jena
Abstract
Reproducibility is one of the fundamental characteristics of science. To reproduce scientific results,
scientists need to manage and describe the provenance of end-to-end experimental pipelines. To un-
derstand, query, and reason how the results are derived, the provenance of the entire study needs to
be described in an interoperable manner. Ontologies play an essential role in representing and inter-
changing provenance information generated in different systems, applications, and domains using a
set of classes, properties, and restrictions. However, ontologies on describing provenance for scientific
studies for different domains have been developed and used in isolation. They should be related to each
other, aligned, and validated to form a network of interlinked ontologies, i.e., an ontology network. To
this end, in this paper, we introduce ReproduceMeON, an ontology network for the reproducibility of sci-
entific studies. The ontology network, which includes the foundational and core ontologies, attempts
to bring together different aspects of the provenance of scientific studies from various applications to
support their reproducibility. We present the development process of ReproduceMeON and the design
methodology of developing core ontologies for the provenance of scientific experiments and machine
learning using a semi-automated approach. We extend our scope to evolve ReproduceMeON to include
ontologies for representing provenance for different subdomains like computational science, bioimag-
ing, and microscopy.
Keywords
Ontology Network, Modeling, Core Ontology, Experiment, Machine Learning
1. Introduction
Reproducibility, the ability to get the same (or close-by) results when repeating an experiment
under different conditions of measurement (e.g., experiment setup, method) [1], is essential
for science as it helps scientists conduct better research in many ways: It allows researchers
to check their results and verify the results of others, thus increasing trust in the scientific
study. It also supports extending and building on top of others’ works, thus promoting scientific
progress. At the same time, achieving the reproducibility of scientific experiments is a complex
real-world problem. Today, scientific studies in large collaborative research projects are often
interdisciplinary and cover data and results from different disciplines. These scientific studies
OntoCom 2021: 8th International Workshop on Ontologies and Conceptual Modeling, held at JOWO 2021: Episode VII
The Bolzano Summer of Knowledge, September 11-18, 2021, Bolzano, Italy
" sheeba.samuel@uni-jena.de (S. Samuel); alsayed.algergawy@uni-jena.de (A. Algergawy);
birgitta.koenig-ries@uni-jena.de (B. König-Ries)
0000-0002-7981-8504 (S. Samuel); 0000-0002-8550-4720 (A. Algergawy); 0000-0002-2382-9722 (B. König-Ries)
© 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073 CEUR Workshop Proceedings (CEUR-WS.org)
involve experiments, their computational environment, wet lab experiments, workflows, com-
putational experiments performed using data science or machine learning (ML) approaches,
etc. Provenance, the source or origin of an object, plays a key role in the reproducibility of
results. It helps understanding the data and sequence of steps performed by scientists, which
led to creating the final result. Hence, researchers need to represent the provenance of results if
they want to report the whole tale of the scientific study. Ontology-based descriptions of the
provenance of data, steps, intermediary, and final results promise to enable reproducibility [2].
Since reproducibility is a complex domain and requirements to describe provenance and meta-
data for different research projects differ in specific aspects, it is not possible to build a large
monolithic domain to cover the entire requirements. Instead, ontologies should be built in
an integrated and modular way, forming a network. An ontology network or a network of
ontologies (short: ON) is defined as a collection of single interconnected ontologies related
to each other via various relationships such as alignment, modularization, and dependency
relationships [3]. In this paper, we claim that ontologies for describing scientific studies for
reproducibility should be organized as ON. We, therefore, introduce the methodology of devel-
oping the ReproduceMe Ontology Network (ReproduceMeON), composed of ontologies that we
have developed and others found in the state of art. For this, we first investigated the state of
the art of ontologies in different relevant areas such as provenance, scientific experiments, ML,
computational, microscopy, and scientific workflows through a systematic literature review.
While the work involved in the development of ReproduceMeON touches upon many topics,
the main focus of this paper is on the introduction of an abstract view of the ON architecture,
where the proposed architecture is based on a three-layer view, including foundational, core,
and domain ontologies. In particular, we propose a ontology matching-based approach to
determine core concepts in each field (e.g., ML, provenance), which can be used later in the core
ontology development. We focus here on how the ontologies have been automatically aligned
and validated. The ontology alignment is done using three systems: OAPT [4, 5], AML [6] and
LogMap [7].
2. Motivation and Use Case
We present a use case scenario showing the experimental workflow of the scientists we in-
terviewed and collaborated with in projects like CRC ReceptorLight [2] and Werkstatt [8]. A
Collaborative Research Center (CRC) is based on a number of interdisciplinary research projects
consisting of several scientists possibly from different disciplines, who work together in teams
towards a common goal. In our case, these include different subdisciplines of biology, medicine
and computer science. The scientific studies conducted by researchers may consist of several
computational and non-computational steps [2]. Figure 1 shows the need for developing an ON
to describe the provenance of scientific studies in an interdisciplinary research project. Scientists
perform wet lab activities like the preparation of samples and solutions, setting up the experi-
ment’s execution environment (room temperature, humidity), etc. These are non-computational
steps which do not directly involve computational resources like computer, software, etc. In
the next steps, several devices like a microscope to capture images of the receptor cell, an
electrophysiological device to generate current, etc., are used in their experiments. The images
Machine Learning
Concepts Ontologies
Feature BigOWL
Algorithm DMOP
Model MEX
Implementation MLSchema
Scientific Experiment ... ... Computational Environment
Concepts Ontologies Concepts Ontologies
Experiment EXPO Function CSO
REPRODUCE- REPRODUCE-
Study Notebook
ME ME
Protocol Script SWO
ISA Interdisciplinary
File Version WICUS
OBI
... Research Project ... ...
...
What are the features of the image generated froma Leica microscope
from a particular experiment that was also used
in machine learning experiment for the detection of tumor cells?
Scientific Workflows BioImaging
Concepts Ontologies Concepts Ontologies
Execution CWL Channel OME Schema
Input D-PROV Image OME-OWL
Port OPMW Microscope REPRODUCE-
Workflow ProvOne Wavelength ME
... ... Provenance ... ...
Concepts Ontologies
Agent OPM
Activity PROV-O
Entity Provenir
Plan P-PLAN
... ...
Figure 1: Motivation for developing an ON. The figure also shows our systematic literature review of
existing ontologies in different areas related to reproducibility
acquired from microscopes are then analyzed by computational tools like proprietary software,
scripts, or Jupyter notebooks based on the complexity of the problem and the skills of scientists.
Reproducing a non-computational step is different from reproducing a computational step. The
provenance of non-computational steps is usually neither machine-controlled nor automatic
and often requires human involvement. Hence, the data, the steps, and the results from the com-
putational and non-computational processes of a scientific experiment need to be interlinked
and described in detail in an interoperable way [2]. In a complete workflow, the process starts
with collecting data generated in the labs and moves on to analyzing and processing them using
several computational techniques like ML.
The REPRODUCE-ME ontology [2] is our first attempt to developing an ontology to describe the
complete path of a scientific experiment consisting of results from the computational and non-
computational steps using semantic web technologies. The complete description, competency
questions used, the development and evaluation of the REPRODUCE-ME ontology are explained
in the paper [2]. It was developed by involving domain experts and computer scientists for the
reproducibility of scientific experiments, initially focusing on the use case of biological imaging
and microscopy [9]. It reuses existing ontologies, PROV-O [10] and P-Plan [11] and also models
the provenance of the execution of scripts and computational notebooks like Jupyter notebooks.
It was used and evaluated in the scientific data management platform CAESAR [2] in the CRC
ReceptorLight project.
Though best practices were used in its development [12] and documentation1 and it fulfills its
1
https://w3id.org/reproduceme/
initial purpose, it was constructed in a monolithic way by providing all the terms based on the
initial use case related to different areas and fields like biological imaging, microscopy, scripts,
computational notebooks, etc. together in one OWL file. Over time, the need to modularize the
REPRODUCE-ME ontology emerged with new requirements. New plans to reuse the ontology
in other projects emerged as it provides the core concepts for describing scientific experiments’
provenance. The reuse of computational provenance described by the ontology is used in
projects like the FAIRification of the PREDICT workflow [13] and intended to be used in data
science in ecological niche modeling [14]. The ontology is also used in computational tools like
ProvBook [15] and ReproduceMeGit [16]. However, to use only its computational provenance
part, currently, the whole ontology has to be imported into these tools and workflows, which in
turn affects the reasoning and performance. The non-computational and computational aspects
of the provenance of scientific studies were described in the ontology without identifying and
separating the modules. Another requirement emerged from the Werkstatt project to describe
the provenance of ML experiments [17]. However, it became challenging to extend the ontology
in its current state. The lessons learned in the development of the REPRODUCE-ME ontology
are used in the development of the ReproduceMeON2 . An ON helps put together under one
umbrella different modules required to describe the provenance of scientific studies.
3. Related Work
Several recent works have developed ON in different domains [18, 19, 20]3 . SEON [19] is a soft-
ware engineering ON which is composed of a foundational ontology, two core ontologies, and
several domain ontologies related to SE subdomains. Their alignment mechanism for integrating
ontologies is by using the ontologies which are grounded in the foundational ontology and
integrating two concepts if they have the same base type. Another recent approach presents de-
veloping an ON in human-computer interaction, HCI-ON [18] which is integrated to SEON [19].
New ontologies are added into the ON and aligned using their own annotation properties.
The motivation behind building these ONs is to organize and structure knowledge in different
domains. Many works have also pointed out the importance of modularizing ontologies [21].
Development of ON and the use of Ontology Design Patterns (ODP) are some of the available
methods in the construction and management of modular and scalable ontologies [3, 21, 22].
Good modular ontologies should have good domain coverage, be formally rigorous, and reuse
foundational ontologies according to [23].
Several ontologies have been developed covering different aspects of the reproducibility of
scientific studies. In prior work [2], we have surveyed different provenance models and on-
tologies covering the computational and non-computational aspects of the reproducibility of
scientific experiments. PROV-O, which provides fundamental concepts for the interoperable
interchange of provenance information among heterogeneous applications and domains, is
widely adopted by the scientific community and is reused and extended by different ontolo-
2
To make it distinct, the REPRODUCE-ME ontology is a single ontology that was developed in [2] to describe
the provenance of scientific experiments focusing on their computational and non-computational aspects. While
the ReproduceMeON is a novel approach evolved from the REPRODUCE-ME ontology and contains a network of
ontologies to describe the provenance of scientific studies.
3
https://github.com/spice-h2020/SON, https://bimerr.iot.linkeddata.es/, https://github.com/rapw3k/glosis
DO1 DO2 DO3 DO4 DO5 DOn
Core Ontologies
Foundational Ontologies
Figure 2: Three-layered ontology network architecture
gies [10]. Provenance ontologies like P-Plan, OPMW, D-PROV, DataONE, ProvONE have been
mainly developed to represent computational processes in scientific workflows and to include
specificities of particular Scientific Workflow Management Systems (SWfMS) [24]. In addition
to the provenance and scientific workflow ontologies, various ontologies have been developed
to capture the provenance of individual domains. The EXPO ontology [25] is developed to
model scientific experiments by describing knowledge about experiment design, methodology,
and results. The Ontology for Biomedical Investigations [26], developed as a community effort
and widely adopted in the biomedical domain, describes experimental metadata in biomedical
research, including planning, execution, and reporting. The recent work [27] presents the OWL
representation of biological imaging data. Few ontologies have also been developed to describe
computational provenance. Software Ontology (SWO) [28] models the data, the version, and
the license used by the software. The REPRODUCE-ME ontology models the provenance of
scientific experiments, bioimaging, scripts, and computational notebooks and their execution [2].
The ReproduceMeON is an initial novel approach to bring different ontologies together for
representing the provenance of scientific studies for their reproducibility. With the development
of the ReproduceMeON, the ability to align and import relevant modules from these ontologies
becomes smooth. The design of ReproduceMeON considers important characteristics like being
modular, considers international standards and reuses foundational ontologies. Our work aims
to implement an ON by applying the characteristics and guidelines for developing modular,
scalable, and reusable ontologies.
4. Development of an Ontology Network
In this section, we introduce the design and development scheme of ReproduceMeON. It is a novel
approach that brings together knowledge from several domains, such as ML, provenance, and
scientific computing, based on the three-layered architecture, as shown in Figure 2. Furthermore,
the proposed approach builds on existing ontologies to enhance knowledge sharing and reuse.
In general, Figure 2 shows that ON is organized into three layers: foundational, core, and
domain-specific ontologies. According to this structure, we have to answer the following
questions:
RQ1 Which are the foundational, core, and domain ontologies that compose the network?
RQ2 Which concepts and relations must be generalized to belong to a core ontology and
specialized to belong to domain-specific ontologies?
RQ3 How should these ontologies in the ON be organized and relate to each other?
In the following, we describe how we can answer these questions.
4.1. Reproducibility Related Area Assimilation
To investigate existing ontologies in the area of reproducibility of scientific studies, we per-
formed a systematic literature review [29]. The need for a systematic review arises from the
requirement to develop an ON for the reproducibility of scientific studies by bringing together
the existing ontologies that have been developed and used by researchers in different domains.
The systematic review answers the research question RQ1. We used Google scholar to identify
the existing ontologies in different areas related to reproducibility from 2006 to 2019. We lim-
ited the search to the following areas: Provenance, Scientific Experiments, Scientific Workflows,
Computational, Machine Learning, and Bio-imaging. Information about the ontologies, includ-
ing the developed year, imported ontologies, documentation, availability, content negotiation,
formalization, and statistics, is available4 .
We found nine ontologies in Provenance, seven ontologies in Scientific Experiments, three in
Bio-imaging, three ontologies in Computational, five ontologies in the ML domain. OPM, PROV-
O, Provenir, P-Plan, OPMW, D-PROV, ProvOne, Research Object Ontology, Common Workflow
Language are the ontologies we found in the area of provenance and scientific workflows. EXPO,
SUMO, OBI, SMART Protocols, Investigation, Study Assay (ISA), The Minimum Information
for Biological and Biomedical Investigation (MIBBI), Bioschemas, REPRODUCE-ME are the
ontologies we found in the area of scientific experiments. MEX Ontology, ML Schema, Prov-ML,
BigOWL, and DMOP are the ontologies we found in the area of ML. Software Ontology (SWO),
WICUS ontology, Function Ontology, REPRODUCE-ME are the ontologies we found in the area
of computational experiments and environment. We found OME Schema, REPRODUCE-ME,
Ontology for an Integrated Image Analysis Platform, and Cellular Microscopy Phenotype On-
tology (CMPO) in the area of bioimaging and microscopy.
We had to exclude some ontologies for the next phase of our study of identifying core ontologies
because of the unavailability of the ontologies. Some are available in a different format other
than the format used in ontologies (e.g., OME Schema is available in XML format). Table 1
shows a snapshot of the ontologies collected using the systematic literature review.
4.2. Core Ontologies Identification
The outcome of the first step is a set of reproducibility-related domains. In each domain, a
number of existing ontologies have been identified and selected to construct the ON. To follow
the three-layered architecture shown in Fig. 2, we need to opt which ontology in each domain
can be used as a core ontology. After that, we build links between the core ontology and the
other ontologies in the same domain (intra-domain links) and then build up links between
ontologies from different domains (inter-domain links).
We started by investigating all collected ontologies, and if there was one that was well-defined
and commonly used as a core ontology in the domain, we selected it as a representative core
4
https://github.com/fusion-jena/ReproduceMeON
Ontology Coverage Serialization Some Concepts
PROV-O Provenance TTL Entity, Activity, Agent, ...
EXPO Experiment OWL ScientificExperiment, ExperimentalTechnology, ...
ISA Experiment OWL Investigation, Study, Assay, ...
SMART Protocol Experiment OWL ExperimentalProtocol, LaboratoryProcedure, ...
REPRODUCE-ME Experiment OWL Experiment, Dataset, Instrument, ...
MEX ML OWL Algorithm, ClassificationProblem, Feature, ...
MLSchema ML TTL Algorithm, Model, Run, ...
DMOP Data Mining OWL ClassificationProblem, DataCharacteristic, ...
OME Schema Microscopy XML Image, ImagingEnvironment, Instrument
Software Ontology Computational OWL License, Software, SoftwareDevelopmentProcess, ...
Table 1
A snapshot of the ontologies collected using the systematic literature review
Figure 3: Core concepts determination
ontology. For example, the PROV-O ontology [10] is widely used in the provenance domain,
providing the foundation to implement provenance applications in different domains, exchange,
and integrate provenance information. Therefore, we selected it as the core ontology for the
provenance part of the ON. For the other parts/domain composing the ON where it is hard to
decide which ontology can be used as a core ontology, we propose an ontology matching-based
approach. The general architecture of the proposed approach is shown in Fig. 3, where the set
of ontologies belonging to each domain are loaded into three different matching tools. Each
matching tool generates a matching result (an alignment) for every pair of ontologies.
• Ontology loading: For each domain, the collected set of ontologies are loaded and pre-
processed (if needed) to be ready for matching, where each ontology pair constitutes
a matching task. i.e., if we have 𝑛 ontologies belonging to a domain, then 𝑛 × (𝑛−1)
2
matching tasks are generated.
• Alignment generation: Ontology matching is a well-know solution to identify similar
entities across a set of different ontologies [30]. We adopt the same idea to determine
intra-domain, and inter-domain links during the development of the ON. Furthermore,
together with a voting algorithm [31], ontology matching can help locate core concepts
in each part/domain of the ON. To this end, we consider three well recognized matching
systems, LogMap[7], AML [6], and OAPT [4, 5], as shown in Fig. 3. We implement a
Figure 4: A mapping example using three matching systems: AML, LogMap, and OAPT
pair-wise matching, where each pair of ontologies from the same domain is loaded into a
matching system constituting a matching task. The corresponding matching result (align-
ment) is generated and saved into a local repository. An alignment is a set of mappings,
usually expressed using the RDF alignment format defined by the ontology matching
community. Each mapping (also called a correspondence), is a quintuple < 𝑖𝑑, 𝑒, 𝑒′ , 𝑐,
𝑟𝑒𝑙 > where: 𝑖𝑑 denotes a unique identifier of the mapping; 𝑒 and 𝑒′ are entities from
two ontologies 𝑂 and 𝑂′ respectively; 𝑐 denotes a measure of confidence, typically a
value within the interval [0, 1], and 𝑟𝑒𝑙 denotes the semantic relation between 𝑒 and 𝑒′
(equivalence (≡), more specific (⊑), more general (⊒), disjunction (⊥)). In the current
implementation, we consider only equivalence (≡) relations.
The mapping example, shown in Fig. 4, illustrates that there is a corresponding between
the "WorkflowTemplate" entity from the BigOwl ontology and the entity "WorkflowTem-
plate" from the DMOP ontology with different confidence values according to the used
matching algorithm. This explains why we consider three different matching tools to
achieve the task, as each tool measures the similarity between ontologies’ entities based
on different aspects.
• Voting: A vote corresponds to the number of times a mapping appeared in the sets
generated by the matching systems. The consensus of vote 2, for instance, will contain
mappings suggested by at least two systems. The more votes, the smaller is the size of
the consensus alignment. We computed the number of mappings produced by applying
Vote 2 and Vote 3 algorithms for scientific experiments and ML. Results are reported in
Tables 2 and 3.
• Alignment validating: Generated alignments are in general not sufficient to be used to
extract core concepts which can be used later for developing the core ontology for many
reasons. First, they allow only a comparison of the systems to each other. Second, they
may contain erroneous mappings, especially if the considered systems use the same
background resources. And finally, valid alignments that have been found by only one
system or none of them will be missing. For this reason, we first validate the generated
Vote 2 Vote 3 Vote 2 Vote 3
No. of mapping 59 14 No. of mapping 82 22
AML 83% 100% AML 93% 100%
LogMap 86% 100% LogMap 37% 100%
OAPT 54% 100% OAPT 96% 100%
Table 2 Table 3
Voting for ML domain Voting for Experiment domain
alignments and add missing concepts based on our knowledge and experience.
Study
Person Sample
p-plan:isSubPlanOfPlan
Entity prov:wasAttributedTo hasMaterial
Measurement
uses hasSetting
xsd:dateTime prov:generatedAtTime, Material
modifiedAtTime
Experiment hasMaterial
p-plan:isVariableOfPlan
Plan p-plan:isSubPlanOfPlan Publication
hasData
Standard
Operating File prov:wasAttributedTo
Procedure
name,
name, description
description
xsd:string xsd:string Author
Figure 5: A portion of schema diagram of the core ontology for scientific experiments. The
orange-filled rectangular box represents the class that the diagram is depicting. The blue-filled
rectangular boxes represent other classes in the ontology. The yellow-filled oval represents a
data type. A subclass relationship is represented by an arrow with a white head and no label. An
arrow with a solid tip represents the relationship mentioned in the label. The class at the solid
tip of the arrow represents the range and the class at the other end of the arrow represents the
domain of the relationship.
• Core concept producing: For each domain, the set of validated alignment is used to extract
concepts that will be used later during the development of the core ontology. We consider
both entities (concepts, relations) from each mapping. For example, the Vote 3 algorithm
generate < 𝑖𝑑, ‘ℎ𝑡𝑡𝑝 : //𝑤𝑤𝑤.𝑘ℎ𝑎𝑜𝑠.𝑢𝑚𝑎.𝑒𝑠/𝑝𝑒𝑟𝑐𝑒𝑝𝑡𝑖𝑜𝑛/𝑏𝑖𝑔𝑜𝑤𝑙𝐴𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚′ , ‘ℎ𝑡𝑡𝑝 :
//𝑤𝑤𝑤.𝑤3.𝑜𝑟𝑔/𝑛𝑠/𝑚𝑙𝑠𝐴𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚′ , 1, ≡>, where the ‘Algorithm’ entities from BigOwl
and MLSchema ontologies. As a result, we consider the term Algorithm as a core term
for the ML domain. ‘CrossValidation’ is another core concept generated using the Vote
3 algorithm from the DMOP and Mex-Core ontologies. Figure 5 shows a portion of
the conceptual model of the core ontology for scientific experiments. The classes are
generated from the validated alignments generated through this step. ‘Study’ is one of the
core concepts validated using the Vote 3 algorithm from the ISA and REPRODUCE-ME
ontologies. Currently, there are 37 core classes identified for scientific experiments and
35 classes for ML through this pipeline.
5. Discussion
We presented an abstract view of the ReproduceMeON and introduced the development of
core ontologies focusing on the area of scientific experiment and ML. However, there are some
open questions in the development of an ON. In the current state of the art, several approaches
exist to align concepts from different ontologies. The most common approach is to import
the entire ontology for the alignment between some concepts of the main ontology. Another
approach is the usage of xref statements [32]. Using own annotation properties to align between
classes from different ontologies is another approach. Each approach has its own pros and cons.
However, importing entire ontology for alignment between concepts of different ontologies can
affect performance and modularity. The ON design should be modular and should not affect
the reasoning power and performance. In our approach of developing the proposed ON, we
need to connect not only the ontologies from same domain but also from different domains.
The question is whether we can use the same approach that we used in linking ontologies in
the same domain to different domains. We also need to design ontologies so that we can easily
plugin different ontologies into the network. The core relationships between the concepts also
need to be identified and generated. Once the core ontologies are developed for the domains
listed in Section 4, the linking of the core ontology to domain ontologies need to be determined.
We plan to do a study on how new ontologies can be integrated into the ON, which are not
aligned with the foundational and core ontologies. We plan to address these open questions in
our future work. We also plan to do an extensive evaluation involving domain experts from
each domain that we have selected for our current ON. The developed ontology will then be
used to apply in the research projects as mentioned in Section 2.
6. Conclusion
This paper presented the need to build and organize ontologies for describing scientific studies
for their reproducibility in an integrated and modular way, forming an ON. We introduced
ReproduceMeON, an ontology network for the reproducibility of scientific studies. As the work
involved in developing an ON is vast, in this paper, we focused on the abstract view of the ON
architecture and the methodology for its development. We conducted a systematic literature
review on the state of the art ontologies in different areas in provenance, scientific experiments,
ML, computational, microscopy, and scientific workflows. We used the result from the review to
develop the proposed ON, which includes foundational, core, and domain-specific ontologies for
representing provenance for different areas like scientific experiments, computational science,
biological imaging and microscopy, and ML. We use ontology matching techniques to select
and develop core ontology for each sub-domain and link to other ontologies in the sub-domain.
In the ON, we plan to build intra-domain and inter-domain links between the core ontologies
and other ontologies from the same and different domains. In addition to the ontologies that are
already integrated or planned to be integrated into the network, we expect ReproduceMeON to
continuously expand by incorporating other provenance ontologies related to the reproducibility
of scientific studies.
Acknowledgments
The authors thank the Carl Zeiss Foundation for the financial support of the project “A Virtual
Werkstatt for Digitization in the Sciences (K3)” within the scope of the program line “Break-
throughs: Exploring Intelligent Systems for Digitization” - explore the basics, use applications”.
Alsayed Algergawy’ work has been funded by the Deutsche Forschungsgemeinschaft (DFG) as
part of CRC 1076 AquaDiva.
References
[1] B. N. Taylor, C. E. Kuyatt, Guidelines for Evaluating and Expressing the Uncertainty of
NIST Measurement Results, Technical Report, NIST Technical Note 1297, 1994.
[2] S. Samuel, A provenance-based semantic approach to support understandability, repro-
ducibility, and reuse of scientific experiments, Ph.D. thesis, University of Jena, Germany,
2019.
[3] P. Haase, S. Rudolph, Y. Wang, S. Brockmans, D1. 1.1 networked ontology model (2006).
[4] A. Algergawy, S. Babalou, M. J. Kargar, S. H. Davarpanah, Seecont: A new seeding-based
clustering approach for ontology matching, in: Advances in Databases and Information
Systems - 19th East European Conference, ADBIS 2015, Poitiers, France, September 8-11,
2015, Proceedings, Springer, 2015, pp. 245–258.
[5] A. Algergawy, S. Babalou, F. Klan, B. König-Ries, Ontology modularization with OAPT, J.
Data Semant. 9 (2020) 53–83.
[6] D. Faria, C. Pesquita, E. Santos, I. F. Cruz, F. M. Couto, Agreementmakerlight 2.0: Towards
efficient large-scale ontology matching, in: Proceedings of the ISWC 2014 Posters &
Demonstrations Track, ISWC 2014, Riva del Garda, Italy, October 21, 2014, volume 1272 of
CEUR Workshop Proceedings, CEUR-WS.org, 2014, pp. 457–460.
[7] E. Jiménez-Ruiz, B. C. Grau, Logmap: Logic-based and scalable ontology matching, in:
The Semantic Web - ISWC 2011 - 10th International Semantic Web Conference, Bonn,
Germany, October 23-27, 2011, Proceedings, Part I, 2011, pp. 273–288.
[8] S. Samuel, M. Shadaydeh, S. Böcker, B. Brügmann, S. F. Bucher, V. Deckert, J. Denzler,
P. Dittrich, F. von Eggeling, D. Güllmar, et al., A virtual “werkstatt” for digitization in the
sciences, Research Ideas and Outcomes 6 (2020).
[9] S. Samuel, B. König-Ries, REPRODUCE-ME: ontology-based data access for reproducibility
of microscopy experiments, in: The Semantic Web: ESWC 2017 Satellite Events - ESWC
2017 Satellite Events, Portorož, 2017, pp. 17–20.
[10] T. Lebo, S. Sahoo, D. McGuinness, K. Belhajjame, J. Cheney, D. Corsar, D. Garijo, S. Soiland-
Reyes, S. Zednik, J. Zhao, PROV-O: The PROV Ontology, W3C Recommendation 30
(2013).
[11] D. Garijo, Y. Gil, Augmenting PROV with plans in P-Plan: scientific processes as linked
data, CEUR Workshop Proceedings, 2012.
[12] A. Gangemi, V. Presutti, Ontology design patterns, in: Handbook on ontologies, Springer,
2009, pp. 221–243.
[13] R. Celebi, J. R. Moreira, A. A. Hassan, S. Ayyar, L. Ridder, T. Kuhn, M. Dumontier, Towards
fair protocols and workflows: the openpredict use case, PeerJ Computer Science 6 (2020)
e281.
[14] M. L. Mondelli, A. Townsend Peterson, L. M. R. Gadelha, Exploring reproducibility and fair
principles in data science using ecological niche modeling as a case study, in: Advances in
Conceptual Modeling, Springer International Publishing, Cham, 2019, pp. 23–33.
[15] S. Samuel, B. König-Ries, Provbook: Provenance-based semantic enrichment of interactive
notebooks for reproducibility, in: Proceedings of the ISWC 2018 Posters & Demonstrations,
Industry and Blue Sky Ideas), USA, 2018, 2018.
[16] S. Samuel, B. König-Ries, Reproducemegit: A visualization tool for analyzing reproducibil-
ity of jupyter notebooks, in: B. Glavic, V. Braganholo, D. Koop (Eds.), Provenance and
Annotation of Data and Processes, Springer International Publishing, Cham, 2021, pp.
201–206.
[17] S. Samuel, F. Löffler, B. König-Ries, Machine learning pipelines: Provenance, reproducibility
and fair data principles, in: Provenance and Annotation of Data and Processes, Springer
International Publishing, Cham, 2021, pp. 226–230.
[18] S. D. Costa, M. P. Barcellos, R. de Almeida Falbo, M. V. H. B. Castro, Towards an ontology
network on human-computer interaction, in: Conceptual Modeling - 39th International
Conference, ER 2020, Vienna, Austria, November 3-6, 2020, Proceedings, volume 12400 of
Lecture Notes in Computer Science, Springer, 2020, pp. 331–341.
[19] F. Borges Ruy, R. de Almeida Falbo, M. Perini Barcellos, S. Dornelas Costa, G. Guizzardi,
Seon: A software engineering ontology network, in: Knowledge Engineering and Knowl-
edge Management, Springer International Publishing, Cham, 2016, pp. 527–542.
[20] C. Sila, O. Belo, V. Barros, Methodology for the development of an ontology network on
the brazilian national system for the evaluation of higher education (ontosinaes), JISTEM-
Journal of Information Systems and Technology Management 15 (2018).
[21] F. Ensan, W. Du, A modular approach to scalable ontology development, in: Canadian
Semantic Web, Springer, 2010, pp. 79–103.
[22] M. C. Suárez-Figueroa, NeOn methodology for building ontology networks: specification,
scheduling and reuse, Ph.D. thesis, Technical University of Madrid, 2012.
[23] M. d’Aquin, A. Gangemi, Is there beauty in ontologies?, Appl. Ontol. 6 (2011) 165–175.
[24] J. Liu, E. Pacitti, P. Valduriez, M. Mattoso, A survey of data-intensive scientific workflow
management, J. Grid Comput. 13 (2015) 457–493.
[25] L. Soldatova, R. King, An ontology of scientific experiments, Journal of the Royal Society In-
terface 3 (2006) 795–803. URL: http://journals.royalsociety.org/content/u552845783800t73/
fulltext.pdf. doi:10.1098/rsif.2006.0134.
[26] R. R. Brinkman, M. Courtot, D. Derom, et al., Modeling biomedical experimental processes
with OBI, J. Biomedical Semantics 1 (2010) S7.
[27] J. Moore, et al., On bringing bioimaging data into the open(-world), 2019. URL: http:
//ceur-ws.org/Vol-2849/paper-06.pdf.
[28] J. Malone, A. Brown, A. L. Lister, J. Ison, D. Hull, H. Parkinson, R. Stevens, The Software
Ontology (SWO): a resource for reproducibility in biomedical data analysis, curation
and digital preservation, Journal of Biomedical Semantics 5 (2014) 25. doi:10.1186/
2041-1480-5-25.
[29] B. Kitchenham, Guidelines for performing Systematic Literature Reviews in Software
Engineering, Technical Report EBSE-2007-01, Keele University and University of Durham,
2007.
[30] P. Ochieng, S. Kyanda, Large-scale ontology matching: State-of-the-art analysis, ACM
Comput. Surv. 51 (2018) 75:1–75:35.
[31] I. Harrow, E. Jiménez-Ruiz, A. Splendiani, M. Romacker, P. Woollard, S. Markel, Y. Alam-
Faruque, M. Koch, J. Malone, A. Waaler, Matching disease and phenotype ontologies in the
ontology alignment evaluation initiative, Journal of biomedical semantics 8 (2017) 1–13.
[32] A. Laadhar, E. Abrahão, C. Jonquet, Investigating one million xrefs in thirthy ontologies
from the obo world, in: 11th International Conference on Biomedical Ontologies, 2020.