Open Science

Practical Issues in Open Research Data

Robert Viseur

1,2

CETIC, Rue des Frères Wright, 29/3, 6041 Charleroi, Belgium

Faculty of Engineering, University of Mons, Rue de Houdain, 9, 7000 Mons, Belgium

Keywords: Open Science, Science 2.0, Open Innovation, Open Source, Open Data, Open Research Data.

Abstract: The term “open data” refers to information that has been made technically and legally available for reuse. In

our research, we focus on the particular case of open research data. We conducted a literature review in

order to determine what are the motivations to release open research data and what are the issues related to

the development of open research data. Our research allowed to identify seven motivations for researchers

to open research data and discuss seven issues. The paper highlights the lack of dedicated data infrastructure

and the need for developing the researcher’s ability to animate online communities.

1 INTRODUCTION

The word “open data” refers to “information that

has been made technically and legally available for

reuse” (Lindman and Tammisto, 2011). Open data

draws interest, because of: the developments in

scientific research (concept of reproducible research

and sharing of experimental data); the enthusiasm,

especially within the scientific community, for the

semantic Web and the linked data; the publications

of datasets in the public sector (e.g. geographical

information); and the emergence of online

communities (e.g. OpenStreetMap). The open data

movement engages the public sector, the businesses

and the researchers. In the scientific community, the

open data adoption is linked to the recent evolutions

in scientific practices. The latter include the online

data sharing and the online sharing of unfinished

works that allow to accelerate the process of

discovery and get feedbacks about the research

(Teif, 2013).

Our paper is dedicated to a particular case of

open data: the open research data. The open research

data falls within the context of so-called science 2.0

and open science. The first one refers to the use of

Web 2.0 practices and tools in the field of scientific

activities. The second one refers to openness.

Practically it can be linked to open innovation and

open source that are popular since a decade. Our

paper is organized in three sections. In the first

section we present the motivations for opening the

research data. In the second section we offer an

inventory of the issues related to the opening of

research data and propose some solutions to address

the issues. In the third section we resume our

findings and discuss the perspectives of our

research.

2 MOTIVATIONS

Some motivations explain the interest in the

scientific community for the open research data.

2.1 Open Access

The open access for the scientific and scholarly

research texts, as defined in the Budapest Open

Access Initiative (www.budapestopenaccess

initiative.org), means “its free availability on the

public internet, permitting any users to read,

download, copy, distribute, print, search, or link to

the full texts of these articles, crawl them for

indexing, pass them as data to software, or use them

for any other lawful purpose, without financial,

legal, or technical barriers other than those

inseparable from gaining access to the internet

itself”. The open access gains popularity through

true open access journals (e.g., PLOS One) and

preprint archives (e.g., ArXiv.org) allowing to

deposit the papers after an embargo period.

201

Viseur R..

Open Science - Practical Issues in Open Research Data.

DOI: 10.5220/0005558802010206

In Proceedings of 4th International Conference on Data Management Technologies and Applications (DATA-2015), pages 201-206

ISBN: 978-989-758-103-8

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

Unfortunately, the words “open” and “open access”

stay to be used with various meanings and may

cause confusion (Murray-Rust, 2008). The

publication of research data with the paper is in line

with the objectives of open access. It is encouraged

by subsequent initiatives such as Panton Principles

for Open Data in Science (pantonprinciples.org).

2.2 Reproducible Research

In computational research the verifiability of

published findings needs the access to data and

computer code. That observation conducts to the

concept of “reproducible research” that would

belong to Claerbout (1992). The latter highlighted

that one of principal goals of scientific publications

was to “provide enough detail to make the work

reproducible”. Many years after his observation that

“in real life, reproducibility is haphazard and

variable”, progress has been slow. For example,

Vandewalle et al., (2009) recently detected failures

of reproducibility in the papers they analyzed in

signal processing field. Their study is based on an

analysis grid to assess the current reproducibility

practices and defines six degrees of reproducibility.

Thus the journal Insight (www.insight-

journal.org) shows how open access, open source

and open data could change scientific practices. It is

an online publication with peer reviewing that is

associated with the software Insight Segmentation

and Registration Toolkit (ITK). The latter is

supported by the company Kitware

(www.kitware.com). ITK is an open source software

tool for image analysis (www.itk.org). The scientific

results are published with the article (as

traditionally) but also with the source code and the

data in order to enhance the reproducibility of the

researches (“reproducible researches”) (Jomier et al.,

2007). The newspaper technical infrastructure

automates the source code compilation and testing.

Other projects such as GNU Octave highlight

those practices (e.g., Eaton, 2012). Several authors

went one step further and developed the concept of

“executable papers” based on the linked data and the

cloud infrastructures (Kauppinen and Espindola,

2011). In addition to publish papers with data and

code, the idea is to provide virtual machines ready to

execute on the cloud or on a local computer without

dealing with dependencies issues.

2.3 Open Innovation

The open innovation paradigm was popularized by

Henry Chesbrough (2006) and treats R&D as an

open system where the company resorts more

broadly to external sources of innovation or to

outlicensing initiatives for its own technologies. The

adaptation of this paradigm to the context of

scientific research is called open science. For

example, the crowdsourcing is applied in scientific

domains.

The crowdsourcing refers to “the outsourcing of

tasks to a crowd that consists of a decentralized,

dispersed group of individuals in a knowledge field

or area of interest” (Schildhauer and Voss, 2014).

The crowdsourcing allows to collectively pool,

aggregate, group and classify data. It may take

different forms. Let see three examples:

DBpedia.org, Observations.be and Amazon

Mechanical Turk (www.mturk.com):

- Dbpedia.org is the core of the semantic Web and

is based on Wikipedia, an online encyclopedia

whose the content is generated by the users

(Auer et al., 2007; Viseur, 2013).

- Observations.be is based on a network of

benevolent passionate skilled lovers of nature to

gather observations related to fauna and flora in

Belgium and Netherlands.

- Amazon Mechanical Turk is a generic

crowdsourcing platform that is frequently used

by scientific community to access panels or

processing more or less significant volumes of

data (Paolacci, 2010).

The collaborative practices are also fostered by the

opportunities offered by the new online tools

inspired by Web 2.0, such as the academic social

networks (e.g., ResearchGate.net) or the online

writing tools (e.g., Google Documents).

2.4 Legal Necessity

Some funding organizations require the researchers

to publish in open access and release open research

data. For example, European Commission define in

his H2020 program a list of domains, called Open

Research Data Pilot, for which a wider access to

publications and data is promoted (EC, 2014). The

decision is motivated by the willingness to improve

the quality of results), to foster collaboration and

avoid duplication of effort, to accelerate innovation

(time-to-market) and involve citizens and society.

2.5 Practical Necessity

Some research fields require the use of large amount

of data. It is for example the case in medicine where

numerous researches have a statistical basis. The

quality and the amount of data become very

DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications

202

important, in particular where the number of tested

factors increases. In consequence, some researchers

build up networks in order to pool the data and

acquire sufficient sample sets (Floca, 2014).

In other disciplines, the researchers face to data

monopolies. Thus the analysis of the World Wide

Web often resorts on commercial search engine due

to the cost of collecting data covering the entire

Web. Unfortunately the researchers face to poorly

documented indexes, secret algorithms and changing

considers “googleology” as a bad science and

proposes an academic-community alternative that

consists in “downloading and indexing substantial

proportions of the World Wide Web, but to do so

transparently, giving reliable figures, and

supporting language researchers’ queries”. Open

data research can thus be a counterweight to data

monopolies.

2.6 Notoriety

Open access can increase the number of citations

(Fecher and Friesike, 2014). Yet citations and

references are the most obvious forms of positive

feedback in the scientific community. Moreover, a

higher number of citations conducts to a greater

likelihood of being quoted in the future, what Fries

(2014) compares to the “Rich get Richer”

phenomenon. Thus, according to Piwowar et al.,

(2007), sharing detailed research data is associated

with increased citation rate, independently of journal

impact factor, date of publication, and author

country.

2.7 Ideology

Fecher and Friesike (2014) identified some schools

promoting a wide access to the product of the

research (e.g., publications and scientific data). The

authors identified five schools (democratic,

pragmatic, infrastructure, public and measurement)

often promoting open data for various reasons such

as the goal to make the knowledge available for

everyone or the improvement of scientific efficiency

through collaboration.

3 ISSUES AND SOLUTIONS

3.1 Traditional Practices

The open data research questions the scientific

practices. Indeed the modern science that came to

life in the 17th century is based on the

professionalization and the institutionalization of the

knowledge creation (Bartling and Friesike, 2014).

The opening of project researches to external world

is a move on traditional practices often based on

individual work or dense collaborations in small

scientific distributed teams. However the increasing

complexity of the problems to be solved needs to

join the efforts and consider multi-expert works in

order to find solutions (Fecher and Friesike, 2014).

3.2 Scientific Publishers Reluctance

The scientific publication is an economic sector

characterized by rising concentration. Thus four big

players control about 60-70% of journal titles

worldwide (Sitek and Bertelmann, 2014) and benefit

on strong position. Some publishers defend their

practices that consist in publicly disseminating the

researches with their source code, the data

structures, the experimental design, the parameters,

the documentation and the figures (Stodden, 2009).

Some researchers try to regain the control of

editorial process and ensure competition between

scientific publishers or alternatively campaign

against the business practices (e.g., “The Cost of

Knowledge” initiative) in order to facilitate the

access to scientific knowledge. The pressure

encounters success as the adoption of open access by

major publishers shows (e.g., Springer Plus).

3.3 Data Property

The publication of open research data require to

choose a license defining the rights and duties of the

licensors and the licensees. Fortunately the Open

Knowledge Foundation (okfn.org) conducts a

project, called Open Data Commons

(opendatacommons.org), aiming to offer set of legal

tools for providing and using open data. They

include three distinct licenses: the Public Domain

Dedication and License, the Attribution License and

the Open Database License (Miller et al., 2008;

Penev et al., 2009). The first one (PDDL) is a public

domain license for data and databases. The second

one (ODC-By) protects the paternity and the third

one (ODC-ODbL) adds a ShareAlike clause that is

similar to copyleft in the free software.

The frequent use of Creative Commons for the

protection of databases is sometimes criticized. The

Belgian and French versions of Creative Commons,

unlike the US version, contain a clause relating to

OpenScience-PracticalIssuesinOpenResearchData

203

the protection of databases. As Creative Commons

licenses are built on copyright, the licensed object

must be a creative work (Miller et al., 2008). Neither

databases nor data are creative work. In addition, the

laws relating to databases differ between the

European Union and the United States. The first

protects databases (see “Directive 96/9/EC”,

europa.eu), but the latter refused to vote on a similar

bill in 1991. However, the use of CC licenses for

data remains recommended by Creative Commons

Foundation herself (e.g., CC0).

The choice of a standard open data license is

recommendable when the data are not critical one

and can be globally published. The situation is more

complex with sensitive data covered by specific

legislations (e.g., private data or medical data).

Solutions such as anonymization techniques are

existing for private data but can be bypassed by

reindentification techniques. The latter are facilitated

by the availability of data that allow to cross

information. The publication may also be hindered if

the data are copyrighted by a third party or protected

under confidentiality agreement.

3.4 University IPR Policies

Universities are gradually encouraged to diversify

their sources of revenues by the sales of licenses

(Intellectual Property Rights), the research contract

signature, the creation of spin-off companies and the

commercial exploitation of new inventions. Geuna

and Nesta (2006) highlights consequences including

a substitution effect between publishing and

patenting for younger researchers and a negative

impact on open science practices in the form of

increased secrecy.

Current public policies in favour of open access

and open research data may redress the balance

between knowledge sharing and business

development in universities.

3.5 Data Semantic

The pooling of data may benefit on features

facilitating the crossing of data. The semantic Web

concept deals with that issue. It leads to the

emergence of Linked Data / Linked Open Data

(LOD) concepts and the creation of several

standards (Bizer, 2009; Miller et al., 2008). Thus the

data related to entities are structured in RDF

(Resource Description Framework). The datasets

can be queried in SPARQL (SPARQL Protocol and

RDF Query Language). DBpedia (dbpedia.org), the

semantic version of Wikipedia, is a concrete

example (Auer et al., 2007; Viseur, 2013).

However, even with semantic standards, the

researcher faces the risk to compare apples and

oranges. The open data research aggregation

supposes to define the semantic of data that were

collected in the linked databases (e.g., metadata) and

the ways the data were collected (e.g, documentation

of heterogeneity over time in chronological series).

3.6 Data Quality

The publication of research results with data and

source code exposes the researchers to a more

detailed review of their studies. The availability of

quality toolkits could foster the researches to publish

their data without risking their reputation.

On the other hand the researchers may benefit on

the data quality to gain positive reputation, as it

works in open source software. Thus the open source

software is considered as a highly individualist

phenomenon that is characterized by a reputation-

based culture (e.g., peer recognition) (Feller and

Fitzgerald, 2002). So talented researchers should

benefit on the visibility offered by high quality open

research data projects.

However, the researchers could benefit on setting

up data quality checking tools on the same model as

software quality. Beyond the specific needs of data

researchers, the development of data quality

standards and norms (e.g., ISO 8000) could help the

structuring of those tools. Researchers may resort on

existing reviews about data quality problems and

tools (e.g., Barateiro and Galhardas, 2005).

3.7 Data Infrastructure

The researchers need public repositories allowing

the publication of their open research data.

We could compare the context with open source

software. If the open source pioneers could accept a

simple Web hosting to publish an archive (e.g., .zip

or .tar files) with their source code, the practices

quickly evolved to dedicated integrated tools that

were called forges (e.g., Sourceforge or Github).

The software forges offered powerful tools in order

to publish, share and concurrently access the source

code for reading or updating (e.g., CVS, SVN or

Git), discuss about the project and document the

defects that were found in the software with bug

tracker (e.g., Bugzilla).

The researchers need the same kind of tool for

open research data. Some tool for general purpose

has been already published (e.g., The Datatank) but

their compliance with researchers requirements must

DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications

204

still be deepened. Indeed the open research data has

its own constraints such as the usability for non

expert users in case of citizen involvement or the

capacity to manage large amount of data in case of

big data projects.

Beyond the software issue the data infrastructure

suppose the ability to fund storage, bandwidth and

processing power over time. Cloud computing offers

on-demand resources but may expose researchers to

lock-in issues, through proprietary interfaces or data

formats (Viseur et al., 2014).

4 CONCLUSIONS

In that research, we identified seven motivations for

researchers to open research data and we also

discussed seven issues (see Table 1).

We assume the researcher’s competences should

evolve to include the ability to create and animate

communities around research projects based on open

source and open data. The benefits are various for

universities and researchers, e.g., citations,

feedbacks or collaborations. Thus the university

technology transfer office (TTO) could support open

source and open data initiatives, as they may

accelerate innovation and contribute to improve the

university image. The help may encompass

community issues, but also IPR (e.g., licenses). The

provision of dissemination tools could be taken in

charge by universities, as they would usefully

complement the already existing institutional

repositories where the affiliate researchers must

deposit their publications, including reports, talks,

conference proceedings or teaching materials.

We highlight an important issue related to the

data infrastructure necessary for sharing data. In

practice some generic tools exist for general open

data initiatives but they are not designed for open

data research. Moreover the open source software

show that the collaborative tools progressively gain

maturity and allow to manage the software source

codes (versioning), the bugs and the documentation.

The most advanced tools allow the continuous

integration and automatically execute some checks

on source codes, resulting in increased software

quality.

Future works should be conducted in order to

identify the tool chain allowing the continuous

integration of open research data (data forge). The

study of projects such as Wikipedia (collaborative

online encyclopaedia) or OpenStreetMap

(collaborative map of the world) should bring

valuable findings. Future data forges should support

the data acquisition (including wrappers, metadata

and semantics), the data storage, the data analysis

(including visualization) and the data export. They

could be based on the results of various research

programs, where executable workflows have already

been settled up for specific purposes (e.g., DataOne;

see (Reichman et al., 2011)), or the existing open

source applications (e.g., Apache Taverna).

Table 1: Summary of the findings.

Motivations

Compliance with open access practices

Needs for reproducible research

Adaptation of open innovation principles to science

Legal necessity

Practical necessity

Search for notoriety

Respect of an Ideology

Issues Solutions

Traditional

practices in

scientific

community

Collaboration as a response to

increasing complexity of the

problems to be solved

Scientific

publishers’

reluctance to open

access to papers and

data

Regain control over editorial

process, pressure on scientific

publishers

Data property

Use of standard licenses for

sharing contents (e.g., Creative

Commons)

University IPR

policies

Setting up public policies

promoting open access and

open research data

Data semantic

Use of semantic Web standards,

use of metadata, documentation

of datasets

Data quality

Automatic testing tools and

their integration with “data

forges”

Data infrastructure

Development of “data forges”

suitable for open research data

REFERENCES

EC (European Commission). Guidelines on Open Access

to Scientific Publications and Research Data in

Horizon 2020 (Version 1.0), 11 December 2013.

Auer, Sören, Bizer, Christian, Kobilarov, Georgi, et al.

Dbpedia: A nucleus for a web of open data. Springer

Berlin Heidelberg, 2007.

Barateiro, José, Galhardas, Helena. A Survey of Data

Quality Tools. Datenbank-Spektrum, 2005, vol. 14, no

15-21, p. 48.

Bartling, Sönke, Friesike, Sascha. Towards another

scientific revolution. In: Opening Science. Springer

OpenScience-PracticalIssuesinOpenResearchData

205

International Publishing, 2014. p. 3-15.

Bizer, Christian. The emerging web of linked data.

Intelligent Systems, IEEE, 2009, vol. 24, no 5, p. 87-

92.

Claerbout, Jon F., Karrenfach, Martin. Electronic

documents give reproducible research a new meaning.

In: 1992 SEG Annual Meeting. Society of Exploration

Geophysicists, 1992.

Chesbrough, Henry, Vanhaverbeke, Wim, West, Joel

(ed.). Open innovation: Researching a new paradigm.

Oxford university press, 2006.

Eaton, John W. Gnu octave and reproducible research.

Journal of Process Control, 2012, vol. 22, no 8, p.

1433-1438.

Fecher, Benedikt, Friesike, Sascha. Open Science: one

term, five schools of thought. In : Opening Science.

Springer International Publishing, 2014. p. 17-47.

Feller, Joseph, Fitzgerald, Brian. Understanding open

source software development. London : Addison-

Wesley, 2002.

Floca, Ralf. Challenges of Open Data in Medical

Research. In : Opening Science. Springer International

Publishing, 2014. p. 297-307.

Geuna, Aldo, Nesta, Lionel JJ. University patenting and its

effects on academic research: The emerging European

evidence. Research Policy, 2006, vol. 35, no 6, p. 790-

807.

Jomier, Julien, Bailly, Adrien, Le Gall, Mikael, et al. An

open-source digital archiving system for medical and

scientific research. Open Repositories, 2010, vol. 7.

Kilgarriff, Adam. Googleology is bad science.

Computational linguistics, 2007, vol. 33, no 1, p. 147-

151.

Kauppinen, Tomi, de Espindola, Giovana Mira. Linked

open science-communicating, sharing and evaluating

data, methods and results for executable papers.

Procedia Computer Science, 2011, vol. 4, p. 726-731.

Lindman, Juho, Tammisto, Yulia. Open Source and Open

Data: Business Perspectives from the Frontline. In :

Open Source Systems: Grounding Research. Springer

Berlin Heidelberg, 2011. p. 330-333.

Miller, Paul, Styles, Rob, Heath, Tom. Open Data

Commons, a License for Open Data. LDOW, 2008,

vol. 369.

Murray-Rust, Peter. Open data in science. Serials Review,

2008, vol. 34, no 1, p. 52-64.

Paolacci, Gabriele, Chandler, Jesse, & Ipeirotis, Panagiotis

G. Running experiments on amazon mechanical turk.

Judgment and Decision making, 2010, vol. 5, no 5, p.

411-419.

Penev, Lyubomir, Sharkey, Michael, Erwin, Terry, et al.

Data publication and dissemination of interactive keys

under the open access model. ZooKeys, 2009, vol. 21,

p. 1-17.

Piwowar, Heather A., Day, Roger S., et Fridsma, Douglas

B. Sharing detailed research data is associated with

increased citation rate. PloS one, 2007, vol. 2, no 3,

e308.

Reichman, O. J., Jones, Matthew B., et Schildhauer, Mark

P. Challenges and opportunities of open data in

ecology. Science, 2011, vol. 331, no 6018.ank-

Spektrum, 14(15-21), 48.

Schildhauer, Thomas, Voss, Hilger. Open Innovation and

Crowdsourcing in the Sciences. In Opening Science.

Springer International Publishing, 2014. p. 255-269.

Sitek, Dagmar, Bertelmann, Roland. Open Access: A State

of the Art. In: Opening Science. Springer

International Publishing, 2014. p. 139-153.

Stodden, Victoria. The legal framework for reproducible

scientific research: Licensing and copyright.

Computing in Science & Engineering, 2009, vol. 11,

no 1, p. 35-40.

Teif, Vladimir B. Science 3.0: Corrections to the Science

2.0 paradigm. arXiv preprint arXiv:1301.2522, 2013.

Vandewalle, Patrick et al. Reproducible research in signal

processing. Signal Processing Magazine, IEEE, 2009,

vol. 26, no 3, p. 37-47.

Viseur, Robert, Charlier, Etienne, Van de Borne Michael.

Cloud Computing and Technological Lock-in:

Literature Review, Data Technologies and

Applications, Vienna, Austria, 2014.

Viseur Robert. Extraction of Biographical Data from

Wikipedia, Data Technologies and Applications,

Reykjavik, Iceland, 2013.

DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications

206