PERICLES – Digital Preservation through Management

of Change in Evolving Ecosystems

Simon Waddington

, Mark Hedges

, Marina Riga

, Panagiotis Mitzias

Efstratios Kontopoulos

, Ioannis Kompatsiaris

, Jean-Yves Vion-Dury

Nikolaos Lagos

Sándor Darányi

, Fabio Corubolo

, Christian Muller

and John McNeill

King’s College London, U.K.

Information Technologies Institute, CERTH, GR-57001 Thessaloniki, Greece

Xerox Research Centre Europe (XRCE), 38240 Meylan, France

Swedish School of Library and Information Science, University of Borås, Borås, Sweden

IPHS, University of Liverpool, L69 3GL, U.K.

B.USOC, Brussels, Belgium

Tate, London, U.K.

{simon.waddington, mark.hedges}@kcl.ac.uk,

{mriga, pmitzias, skontopo, ikom}@iti.gr,

{Jean-Yves.Vion-Dury, Nikolaos.Lagos}@xrce.xerox.com,

Sándor.Darányi@hb.se, corubolo@gmail.com,

christian.muller@busoc.be, john.mcneill@tate.org.uk

Abstract. Management of change is essential to ensure the long-term reusabil-

ity of digital assets. Change can be brought about in many ways, including

through technological, user community and policy factors. Motivated by case

studies in space science and time-based media, we consider the impact of

change on complex digital objects comprising multiple interdependent entities,

such as files, software and documentation. Our approach is based on modelling

of digital ecosystems, in which abstract representations are used to assess risks

to sustainability and support tasks such as appraisal. The paper is based on

work of the EU FP7 PERICLES project on digital preservation, and presents

some general concepts as well as a description of selected research areas under

investigation by the project.

1 Introduction

1.1 Motivation

Existing approaches to digital preservation are heavily influenced by practices that

have evolved over many years in the non-digital world. The reusability of digital ob-

jects is dependent on their surrounding environment. This can include not only rele-

vant software, but also platforms and documentation, and often the digital objects and

their environment have complex interdependencies. Due to the rapid pace of techno-

logical change, the environment in which a digital object exists will evolve and some

entities may become delinked or even obsolete. This may result in the loss of capabil-

Muller C., Vion-Dury J., Lagos N., Kontopoulos E., Riga M., Mitzias P., Corubolo F., Hedges M., DarÃ ˛anyi S., Waddington S., Kompatsiaris Y. and McNeill J.

PERICLES â

A ¸S Digital Preservation through Management of Change in Evolving Ecosystems.

DOI: 10.5220/0006163600510074

In The Success of European Projects using New Information and Communication Technologies (EPS Colmar 2015), pages 51-74

ISBN: 978-989-758-176-2

ity to run software, to interpret information or to render data files. A similar argument

can be applied to other types of change. For example, organisational changes may

result in digital objects held by the organisation no longer being compliant with cur-

rent policies and procedures. Evolution of user communities may result in the digital

objects being interpreted by individuals and used for purposes that were not envisaged

when they were initially created or acquired. This can result in the digital objects not

being fit for purpose or even understandable by current users.

Maintaining digital objects in a static form in a repository, as might be done with

non-digital assets such as books and paintings, which can remain in a reusable form

for centuries, is unlikely to be successful even over time periods of a few years. Thus

new approaches are required to managing such digital objects that can deal with both

complex dependencies as well as continual change.

1.2 PERICLES Objectives and Approach

The main challenge for PERICLES is to ensure the ongoing interpretation and reusa-

bility of digital objects that are heterogeneous, volatile (i.e. subject to continual

change) and are complex (i.e. have many interdependencies). By analogy with biolog-

ical systems, we use the term digital ecosystem to reflect an evolving set of interde-

pendent entities, which is subject to influences bringing about change. Digital ecosys-

tems can include any entities that can have a direct or indirect impact on the reuse of

digital objects, including data objects, software, user communities, processes, tech-

nical services and policies. An important feature of our approach is that a definition of

a digital ecosystem includes descriptions of the dependencies between the constituent

entities.

Following a widely adopted methodology in science, we introduce computational

models to enable the impact of change on a digital ecosystem to be assessed, and in

some cases for mitigating actions to be determined, without the need to manipulate the

entities in the ecosystem directly. Based on a linked data paradigm, the models make

use where possible of existing domain ontologies. The evolution of the models is

governed by policies.

In order to populate such models, tools are provided to support the extraction of

metadata, such as content, environmental, usage and provenance information. Analysis

and visualisation of the models is used to support risk analysis and provide decision

support. Finally preservation actions can be determined and translated into executable

business processes.

To support the development, testing and real-world deployment of PERICLES

components, an integration framework is under development. This includes an Entity

Registry and Model Repository to support the storage and retrieval of the models as

well as an execution layer to enable preservation components to be wrapped in han-

dlers and run against the stored entities. The integration framework also provides a

reference implementation for deploying PERICLES components in real-world applica-

tions.

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

1.3 Digital Preservation Activities in the EU

In this section, we briefly review prior EU-funded activities relating to digital preser-

vation, to place PERICLES in a wider context. In 2001, the EU funded the Electronic

Resource Preservation and Access Network (ERPANET) project

in the FP5 pro-

gramme. This was the first attempt to engage with memory organisations, such as

museums and libraries, and commercial sector for the purpose of raising awareness

about the need for digital preservation and at the same time providing the necessary

knowledge base to all participants.

There then followed approximately €100M of EU funding for digital preservation,

covering a wide range of topics and including research and development prototypes.

The PLANETS

project developed the Planets Suite, comprising a preservation plan-

ning tool, a test-bed and an interoperability framework. The planning tool, Plato, of-

fered information on digital objects at risk, and supported informed decision-making

on preservation actions. The project primarily dealt with simple digital objects, rather

than complex dependencies, and used a sampling approach on individual objects to

evaluate preservation actions.

CASPAR

worked primarily on preservation approaches to validate the OAIS ref-

erence model [13] in the cultural, artistic and scientific domains. It investigated the

implementation and use of key OAIS concepts such as representation information,

knowledge management and preservation description information. SHAMAN

studied

the incorporation of Product Lifecycle Management within a digital preservation sys-

tem, and produced an extended information lifecycle model. PROTAGE

explored the

use of software agents targeting automation of digital preservation processes. The

LiWA

project dealt with archiving of web content.

TIMBUS

, addressed preservation of business processes where software and plat-

form are developed and delivered as a service. SCAPE

focused on scalable preserva-

tion algorithms, extending the results of PLANETS to high volume content. The

ENSURE

project considered scalable pay-as-you-go infrastructure for preservation

services based on cloud computing technology, as well as exploring non-traditional

domains for digital preservation such as finance and medicine. Finally the

APARSEN

network tried to join together work on pervious preservation projects

into a common vision, again underpinned by the OAIS model.

PERICLES differs from most preceding projects in that it considers continuously

changing environments such as for time-based media, where OAIS is less appropriate.

Our approach is based on a continuum viewpoint. Although static dependency models

http://www.erpanet.org/index.php.

http://www.planets-project.eu/

http://shaman-ip.eu/

http://www.ra.ee/protage

http://liwa-project.eu/

http://timbusproject.net/

http://www.scape-project.eu/

http://ensure-fp7-plone.fe.up.pt/site

http://www.alliancepermanentaccess.org/index.php/aparsen/

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

were considered in earlier projects, such as CASPAR, the use of dynamic models is

new. There has also been relatively little work done to date on semantic change, which

is an important focus of PERICLES.

1.4 Acknowledgements

This work was supported by the European Commission Seventh Framework Pro-

gramme under Grant Agreement Number FP7-601138 PERICLES. The authors wish

to acknowledge the contributions of our many PERICLES colleagues to the content of

this paper.

2 Digital Preservation and Change

2.1 Lifecycle versus Continuum Approaches to Digital Preservation

Lifecycle models are a point of reference for most existing approaches and practices

in digital preservation. They provide a framework for describing a sequence of actions

or phases, such as creation, productive use, modification and disposal, for the man-

agement of digital objects throughout their existence. Such models suggest a linear

sequence of distinct phases and activities, which in practice may be non-linear or even

relatively disordered. Lifecycle models provide an idealised abstraction of reality, and

may typically be used in higher-level organisational planning and for detecting gaps in

procedures.

The DCC lifecycle model [10] is one of the most well-known lifecycle models. It

provides a graphical, high-level overview of the stages required for successful cura-

tion and preservation of data from initial conceptualisation or receipt through the

iterative curation cycle. The UK Data Archive describes a research data lifecycle

which comprises six sequential activities and, unlike the DCC model, is more focused

on the data user’s perspective. Overviews of lifecycle models for research data are

provided by Ball [11] and the CEOS Working Group on Data Life Cycle Models and

Concepts [12].

So-called lifecycle approaches typically envisage a clear distinction between active

life and post-active life. The Open Archival Information System (OAIS) [3] is a com-

monly adopted reference model for an archive, consisting of an organisation of people

and systems that has accepted the responsibility to preserve information and make it

available for a designated community. Although lifecycle models and OAIS provide a

useful frame of reference for preservation, they are less suited to dealing with exam-

ples where there is a less clear distinction between the active life and archival phases,

examples of which will be discussed in section 3.

In [2], we introduced a continuum approach to digital preservation that combines

two main aspects. Firstly, there is no distinction made between active life and post-

active life; that is, preservation is fully integrated into the active life of the digital

http://ijdc.net/index.php/ijdc/article/view/69

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

objects. A second aspect is that preservation is non-custodial, that is we do not aim

necessarily to remove entities from their environment, both physical and organisation-

al, and place them in the custody of a third party.

Continuum approaches have been proposed in the closely related field of record

keeping. The Records Continuum (RC) was originally proposed by Upward in 1996

[4]. An essential aspect is that the content and structure of a record are fixed, but the

surrounding context can change over time, so a record is “always in a state of becom-

ing” [15].

2.2 Change Types and Their Impact

PERICLES considers a number of different types of change that can potentially have

an impact on the reuse of digital objects. A more extensive review is presented in [16].

The main high-level change types that we have considered are summarised in Table 1.

Table 1. Types of change occurring on digital ecosystems and their impact.

Type of change

Description

Impact

Knowledge and

terminology

Changes in semantics that originate

from a designated user community.

Different user communities

using the same underlying

datasets with different under-

standing and goals.

Technology

This includes hardware availability,

software obsolescence, and changes in

formats, protocols and interfaces.

Requires replacement of hard-

ware and software compo-

nents, transcoding of files,

redesign of interfaces etc.

Policy

Changes in permissions, legal re-

quirements, quality assurance and

strategy.

This can impact how and

where digital objects are

stored, quality processes they

are subjected to, retention

periods etc.

Organisation

Change to the organisation due for

example to political, financial or stra-

tegic reasons. Often organisational

changes can be manifested as policy

changes.

This can result in different

priorities for retaining or main-

taining the reusability of digi-

tal objects.

Practice

This change originates from new or

changed habits of the designated user

community (not necessary related to

knowledge and terminology changes).

It is an indicator that user requirements

may change.

This can result in changes to

the form in which digital ob-

jects are retained, reflecting the

changing ways in which they

are to be reused.

Requirements

This can include business require-

ments, functional requirements that a

system should fulfil, quality of service

and user requirements.

This again reflects the way that

digital objects are reused and

hence how they should be

stored and maintained.

Dependency

Either characteristic attributes of a

dependency are changed (e.g. quicker,

faster, more flexible, cheaper) or the

dependency itself changes.

Evolution in dependencies can

reflect different views on the

types of change that are being

considered.

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

2.3 Change and Dependency

Change and dependency can in many respects be viewed as dual notions. Thus the

types of dependency we may wish to model are related to the types of change that are

being addressed. In PERICLES, we say that entity A is dependent on entity B if

changes to B have a significant impact on the state of A. A key aspect of PERICLES

is that dependencies can have associated semantics and do not merely represent a link

between the two objects. The semantics of a dependency are related to the change

context under consideration.

A number of notions of dependency exist in the literature. The PREMIS Data Dic-

tionary

defines three types of relationships between objects: structural, derivation

and dependency. In particular, a derivation relationship results from the replication or

transformation of an object. A dependency relationship exists when one object re-

quires another to support its function, delivery, or coherence.

The Open Provenance Model (OPM)

introduces the concept of a provenance

graph that aims to capture the causal dependencies between entities. The most relevant

concept from our perspective is process that represents actions performed on or

caused by artefacts, and resulting in new artefacts.

In a preservation context, [17] defines notions of module, dependency and profile

to model use by a community of users. A module is defined to be a software/hardware

component or knowledge base that is to be preserved, and a profile is the set of mod-

ules that are assumed to be known to the users. A dependency relation is then defined

by the statement that module A depends on module B if A cannot function without B.

For example, a README.txt file depends on the availability of a text editor (e.g.

Notepad). The authors of [8] also define the more specific notion of task-based de-

pendency, expressed as Datalog rules and facts. In [19], the notion of task is extended

to intelligibility, which allows for typing dependencies. The PERICLES modelling

approach goes one step further toward genericity, by allowing any kind of dependency

specialisation, and provides a much richer topology for dependency graphs through

managing dependencies as objects instead of properties.

2.4 Semantic Change

An important aspect of PERICLES is the study of evolving semantics and semantic

change in particular. The risk of semantic change for digital preservation is that, as a

fallout from inevitable language evolution that has been accelerating due to an inter-

play of factors, future users may lose access to content, either (a) because the concepts

and/or the words as their labels will have changed, or (b) because the same concept

may have different labels over separate user communities.

Additionally, by better understanding semantic change processes, one can identify

‘at risk’ terminology, which is likely to be hard to understand by future users of the

resources. Similarly, one can identify specialist terminology likely to be different

across domains and, therefore, difficult for those other domain users to understand.

http://www.loc.gov/standards/premis/

http://eprints.soton.ac.uk/271449/ 1/opm.pdf

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

Another issue for semantic change might relate to the change over time in the way in

which a resource is used.

In the investigation of semantic change, PERICLES is conducting experiments in

two major directions: one looking at the interpretability of digital objects over time

and over designated user communities and the other focusing on drift

detection,

measurement and quantification methods. The aim is to eventually relate the two in a

common frame of thought.

When considering drift interpretation (understandability), one of the key questions

is how to represent the knowledge model of a domain, community, or individual. This

is a critical question as any subsequent analysis or experimentation will depend upon

the quality and ‘soundness’ of any methods or assumptions made at this initial stage.

Regarding drift quantification (measurability), vector- versus graph-based measures

are computed and ranked. In this way, all kinds of shifts could be analysed, be they

community-dependent or temporal.

A key element of our explorations is to address drifts in word meaning used for

document indexing, and consequent changes in document meaning together with topic

shifts typical of evolving document collections. To this end, we also call in probabilis-

tic methods such as additive regularisation based topic models [20]. Secondly, as in

classical mechanics, physical systems in change are typically analysed by calculus and

represented as vector fields. Using physics as a metaphor we introduced a vector field

based tool to study evolving semantics [21] and its scalability aspects [22]. This work

is in the phase of adding qualitative evaluation of shifts in word meaning to the model,

which is novel because typically, only quantitative drifts have been addressed by

measurements [24]. Finally, our vector field model of semantic change points in the

direction of social mechanics [25, 26], thereby paving the way for an integrative meta-

theory of changes in sign systems as a function of social use depending on evolving

sign contexts.

3 Case Studies

The examples selected for study in PERICLES are chosen from the application do-

mains that the project is addressing, namely digital media, and space science.

3.1 Examples from the Digital Media Domain

Within the PERICLES project, the digital media domain covers three different sub-

domains, namely Digital Video Art (DVA), Software-Based Art (SBA) and Born-

Digital Archives (BDA). Several key challenges have been defined within each of

these subdomains and corresponding ontologies have been developed; these do not

attempt to model the respective subdomains exhaustively, but are primarily aimed at

modelling preservation-related risks. Specifically, in DVA, the focus is on the con-

sistent playback of digital video files, with respect to the technical or conceptual char-

In literature, semantic change is covered by expressions like semantic drift, semantic shift,

semantic decay, and sometimes as concept drift.

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

acteristics of the corresponding digital components. In SBA, the focus is on the as-

sessment of risks for newly acquired artworks, regarding their technical dependencies

but also the functional, conceptual and aesthetic intentions related to the significant

properties of an artwork. Finally, within the BDA context, the focus is on the need to

be able to access and maintain digital documents as was originally intended, together

with all their technical, aesthetic and permission characteristics. A detailed analysis of

the media case study is contained in [27].

All key challenges are explored in relation to collections held by Tate

. The DVA

and SBA collections belong to the main art collection, while the BDA material exists

in the Tate Library and Archive collection; these two types of collections are managed

by different teams within Tate. The institution has approximately 300 video artworks,

including digital video artworks, in the collection. It also has a small but growing

number of software-based artworks. The born-digital material in Tate Library and

Archive includes material from institutions within the UK, such as records from com-

mercial galleries that come into the archive, as well as artists’ personal records. Much

of this material comprises standard formats such as emails, spreadsheets, text docu-

ments, images, and so on.

3.2 Examples from the Science Domain

B.USOC

supports experiments on the International Space Station (ISS) and is the

curator of both collected data and operation history. B.USOC chose to analyse the

SOLAR payload, in operation since 2008 on the ESA COLUMBUS module of the

ISS. These observation data are prime candidates for long-term data preservation, as

variabilities of the solar spectral irradiance have an influence on Earth’s climate, and

the measurements cannot be repeated. The current SOLAR module is built from three

complementary space science instruments (see Fig. 1) that measure the solar spectral

irradiance with an unprecedented accuracy.

Fig. 1. The International Space Station, and the SOLAR module as part of the COLUMBUS

(left) and the SOLAR instrument (right).

http://www.tate.org.uk/

http://www.busoc.be

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

SOLAR and the phenomena it studies are an example of change in observation da-

ta with the evolution of knowledge on the sun. Before the space age (forty years ago),

the sun’s input to the earth system was called the “solar constant” and great pains were

taken to determine it by removing the atmospheric effects. Then in the beginning of

the 1980’s, several instruments, including first versions of SOLAR, were flown in

space and determined that the “constant” was in fact a variable parameter synchronous

with the 11 year sunspot cycle, hence its name was changed to “total solar irradiance”.

Moreover, spectral variations were found not be uniform and that the ultraviolet re-

gion, while weak in energy, had important variations relating to solar activity known

now as space weather (solar flares and other phenomena).

The SOLAR instruments, which had been designed originally to provide snapshots

of the solar spectral irradiance at well-defined parts of the solar cycle, now deliver

valuable scientific data relevant to both shorter and longer time scales.

From a digital preservation perspective, the experiments consist of highly complex

interlinked digital entities, including raw data and associated telemetry, software,

documentation (over a hundred document categories), and operational logs.

4 Model-driven Approach

4.1 Functional Architecture

Following a common paradigm in science, we introduce models to enable the impact

of change on a digital ecosystem to be assessed, and in some cases for mitigating

actions to be determined. This principle is illustrated in Fig.2, which describes the

PERICLES functional architecture, based on [5].

Fig. 2. The PERICLES functional architecture.

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

The components in blue, joined by solid lines, represent the main workflow for

building the models, performing change impact analysis, determining preservation

processes or actions, validating the results of the preservation actions, and updating

the models. The remaining components support the model-driven workflow. A user

interface component is required both to create and in some cases populate the models,

as well as to perform change impact analysis and determination of preservation ac-

tions. The Entity Registry/Model Repository (ERMR) assigns unique identifiers to the

entities in an ecosystem and provides tools for storing and retrieving the models them-

selves. The Knowledge Base provides the underlying ontologies for constructing the

ecosystem models, to be described in the following section, together with reasoning

tools. Finally the preservation actions are executed via a workflow engine, using a set

of transformation components.

4.2 Digital Ecosystem Models

The PERICLES Linked Resource Model (LRM) is an upper level ontology

designed to

provide a principled way to modelling evolving ecosystems, focusing on aspects relat-

ed to the changes taking place. This means that, in addition to existing preservation

models that aim to capture provenance and preservation actions, the LRM also aims at

modelling how potential changes to the ecosystem, and their impact, can be captured.

It is important to note here that we assume that a policy governs at all times the dy-

namic aspects related to changes (e.g. conditions required for a change to happen

and/or impact of changes). As a consequence, the properties of the LRM are depend-

ent on the policy being applied. At its core the LRM defines the ecosystem by means

of constituent entities and dependencies. The main concepts of the static LRM are

illustrated in Fig. 3. (The prefix pk refers to the LRM namespace).

Resource. Represents any physical, digital, conceptual, or other kind of entity; entities

may be real or imaginary and in general comprises all things in the universe of dis-

course of the LRM Model. A resource can be Abstract (c.f. AbstractResource in

Fig. 3), representing the abstract part of a resource, for instance the idea or concept of

an artwork, or Concrete (c.f. ConcreteResource), representing the part of an entity

that has a physical extension and is therefore characterized by a location attribute

(specifying some spatial information). These two concepts can be used together to

describe a resource; for example, both the very idea of an artwork, as referred by

papers talking about the artist’s intention behind the created object, and the corre-

sponding video stream that one can load and play in order to manifest and perceive the

artwork. To achieve that, the abstract and concrete resources can be related through a

specific realizedAs predicate, which in the above example could be used to express

that the video file is a concrete realization of the abstract art piece.

Dependency. An LRM Dependency describes the context under which change in one

or more entities has an impact on other entities of the ecosystem. The description of a

dependency minimally includes the intent or purpose related to the corresponding

usage of the involved entities. From a functional perspective, dedicated policies/rules

further refine the context (e.g. conditions, time constraints, impact) under which

change is to be interpreted for a given type of dependency.

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

Fig. 3. Main concepts of the static LRM.

For example, consider a document containing a set of diagrams that has been creat-

ed using MS Visio 2000, and that a corresponding policy defines that MS Visio draw-

ings should be periodically backed up as JPEG objects by the work group who created

the set of diagrams in the first place

. According to the policy, the work group who

created the set of JPEG objects should be able to access but not edit the corresponding

objects. The classes and properties related to Dependency can be used to describe

each such conversion in terms of its temporal information and the entities it involves

along with their roles in the relationship (i.e. person making the conversion and object

being converted), as other existing models. In addition, the LRM Dependency is

strictly connected to the intention underlying a specific change. In the case described

here the intent may be described as “The work group who created the set of diagrams

wants to be able to access (but not edit) the diagrams created using MS Visio 2000.

Therefore, the work group has decided to convert these diagrams to JPEG format”

and it implies the following.

 There is an explicit dependency between the MS Visio and JPEG objects. More

specifically, the JPEG objects are depending on the MS Visio ones. This means

that if an MS Visio object ‘MS1’ is converted to a JPEG object, ‘JPEG1’, and

‘MS1’ is edited, then ‘JPEG1’ should either be updated accordingly or another

JPEG object ‘JPEG2’ should be generated and ‘JPEG1’ optionally deleted (the de-

scription is not explicit enough here to decide which of the two actions should be

performed). This dependency would be especially useful in a scenario where MS

Visio keeps on being used for some time in parallel to the JPEG entities being used

as back up.

 The dependency between ‘MS1’ and ‘JPEG1’ is unidirectional. Actually, JPEG

objects are not allowed to be edited and, if they are, no change to the corresponding

MS Visio objects should apply.

 The dependency applies to the specific work group, which means that if a person

from another work group modifies one of the MS Visio objects, no specific conver-

This example is adapted from a use case described in [19], pp. 52-53.

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

sion action has to be taken (the action should be defined by the corresponding poli-

cy).

To enable recording the intent of a dependency, we can relate in the LRM the De-

pendency entity with an entity that describes the intent via a property that we name

intention, as illustrated in Fig. 4.

Fig. 4. A view of the Dependency concept in LRM.

Let us take once more the example above: we need to be able to express the fact

that a transformation to the JPEG is possible only if the corresponding MS Visio ob-

ject exists and if the human that triggers the conversion has the required permissions

to do that (i.e. belongs to the specific workgroup). The impact of the conversion (gen-

erating a new JPEG object) could also be conditioned on the existence of a corre-

sponding JPEG object containing an older version of the MS Visio object. The actions

to be undertaken in that case, would be decided based on the policy governing the

specific operation. Assuming that only the most recent JPEG object must be archived,

the old one must be deleted and replaced by the new one (conversely deciding to keep

the old JPEG object may imply having to archive the old version of the corresponding

old MS Visio object as well).

Plan. The condition(s) and impact(s) of a change operation are connected to the De-

pendency concept in LRM via precondition and impact properties as illustrated

in Fig. 4. These connect a Dependency to a Plan, which is defined as a specialized

description representing a set of actions or steps to be executed by some-

one/something (either human or software); this is, thus, a means to give operational

semantics to dependencies. Plans can describe how preconditions and impacts are

checked and implemented (this could be for example defined via a formal rule-based

language, such as SWRL). The temporally coordinated execution of plans can be

modelled via activities. A corresponding Activity class is defined in LRM, which

has a temporal extension (i.e. has a start and/or end time, or a duration). Finally, a

resource that performs an activity, i.e. is the “bearer” of change in the ecosystem,

either human or man-made (e.g. software), is represented by a class called Agent.

4.3 Domain Ontologies

A domain ontology (or domain-specific ontology) is a formal description of modelling

concepts in a specific domain in a structured manner. Three media domain ontologies

have been developed within PERICLES, aimed at modelling digital preservation risks

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

for the three respective subdomains (DVA, SBA and BDA), via LRM-based con-

structs (c.f. section 3.1). The key notions adopted and extended by the LRM are:

 Activity - represents activities that may be executed during a digital item’s

lifespan. The media domain ontologies extend the Activity class, in order to mod-

el activities that are considered to be important for digital preservation processes

(creation, acquisition, storage, access, display, copy, maintenance, loan, destruc-

tion, etc.).

 Agent - with subclasses for human and software agents. Human agents are in

addition specialised for the media domain into artists, creators, programmers, mu-

seum staff, etc. and software agents into programs, software libraries, operating

systems, etc.

 Dependency (c.f. Section 4.2) - indicates the association or interaction of two or

more resources in the domain ontology. For the media domain ontologies, we ex-

tend the basic notion of LRM dependency with three sub-categories:

o Hardware Dependencies - specify hardware requirements for a Re-

source in order for it to function properly.

o Software Dependencies - indicate the dependency of a Resource or Ac-

tivity on a specific software (Software Agent) - name, version, etc.

o Data Dependencies - imply the requirement of some knowledge, or data

or information, in order for a Resource to achieve its purpose of exist-

ence or function. This kind of data may originate from human input (e.g.

passwords), computer files (e.g. configuration files), network connec-

tion, live video, etc.

The context of dependencies may be additionally enriched with the notions of in-

tention and specification (see Section 4.2). For the media domain ontologies, a set

of predefined intention types were defined:

o Dependencies with a Conceptual Intention are aimed at modelling the

intended “meaning” of a resource (i.e. artwork) by its creator; according

to the way he/she meant it to be interpreted/understood. For example, a

poem (digital item) belonging to an archival record may not conserve its

formatting during the normalization process, something that may be

contrary to the intention of the poet regarding the way that the poem is

conceptualised/conceived by a reader.

o Dependencies with a Functional Intention represent relations relevant

to the proper, consistent and complete functioning of the resource. For

example, a specific codec is required to display a digital video artwork.

o Dependencies with a Compatibility Intention model compatible soft-

ware or hardware components which may operate together or as re-

placement components for availability, obsolescence or other reasons.

For example, the software used for playing back a digital video artwork

consistently is compatible with certain operating systems.

A domain-specific instantiation, presented in Fig. 5, describes the following sce-

nario taken from the BDA subdomain: a normalisation activity is applied in a text file

(item, digital resource), according to the archival policy defined by the used normali-

sation software (OpenOffice). In terms of the media ontology, there is (a) a data de-

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

pendency of the normalisation activity to the item itself, (b) a software dependency of

the normalization activity to the used software, and (c) a hardware dependency of the

normalization software to the hardware required in order for the software to run effi-

ciently. The intention of all three types of dependencies is functional, meaning that all

the required resources modelled in this example impact the functionality of the re-

sources for which the dependencies were implemented.

Fig. 5. Dependencies existing within the context of a normalisation activity applied in a digital

item.

Within the context of PERICLES and the DVA ontology, an Ontology Design Pat-

tern (ODP) for representing digital video resources was introduced [6]. This work was

motivated by the problem of consistent presentation of digital video files in the con-

text of digital preservation. The aim of this pattern is to model digital video files, their

components and other associated entities, such as codecs and containers (Fig. 6). The

proposed design pattern facilitates the creation of relevant domain ontologies that will

be deployed in the fields of media archiving and digital preservation of videos and

video artworks.

Fig. 6. Digital Video ODP schematic view.

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

The design pattern illustrates a more general principle, namely that ecosystem

models can be constructed from a set of common templates. Such an approach would

greatly reduce the effort required to create models by enabling a more modular ap-

proach where templates are reused across many different models.

4.4 Environment Capture and Model Population

An important consideration for a model-driven approach to preservation is to mini-

mise the effort required to construct the ecosystem models. The PERICLES Extraction

Tool (PET) provides one approach. PET is an open source framework for the extrac-

tion of the Significant Environment Information of a digital object. Here significance

is a positive number expressing the importance of a piece of environment information

for a given purpose. The tool can be used in a sheer curation [23] scenario, where it

runs in the system background and reacts to events related to the creation and altera-

tion of digital objects and the information accessed by processes, to extract environ-

ment information with regard to these events. All changes and successive extractions

are stored locally on the curated machine for further analysis. A further extraction

mode is the capturing of an environment information snapshot, which is intended for

the extraction of information that does not change frequently.

The tool aims to be generic as it is not created with a single user community or use

case in mind, but can be specialised with domain specific modules and configuration.

PET provides several methods for the extraction of SEI, implemented as extraction

modules as displayed in Fig. 7. The configuration has to be done once, after that PET

can run automatically in a way that does not interfere with system activities and fol-

lows the sheer curation principles. The PET tool can be used to generate models based

on the LRM ontology, which can then be used for ecosystem analysis. PET has been

released as open source software under the Apache license. PET source code together

with documentation and tutorials are available for download on Github

Fig.7. SEI extraction with the PERICLES Extraction Tool.

https://github.com/pericles-project/pet

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

5 Applications in Digital Preservation

5.1 Monitoring of User Communities

One complex issue facing PERICLES is tracking the use of media (images or video)

by different user communities. Although it is possible to apply usage tracking to a web

portal making archival content available, once content is downloaded its use can no

longer be monitored. Even this presupposes that portal users are registered and pro-

vide details of their intended usage, which may also not be feasible. What would be

helpful in this context is the possibility to embed metadata and identifiers that would

allow mapping and monitoring the diffusion of the media across user communities, in

order to help identify their evolution.

Information encapsulation (IE) methods can be distinguished into the categories of

packaging and information embedding. Packaging refers to the aggregation of files or

other information formats as equal entities stored in an information container. In con-

trast to this, information embedding needs a carrier information entity (file/stream) in

which payload information will be embedded.

The PERICLES Content Aggregation Tool (PeriCAT) [28] is a framework for In-

formation Encapsulation techniques. It integrates a set of information encapsulation

techniques from various domains, which can be used from within the framework.

Furthermore PeriCAT provides a mechanism to capture the scenario of the user, and

to suggest the best fitting information encapsulation technique for a given scenario.

The tool is available to download on Github

under an Apache Version 2.0 licence.

5.2 Quality Assurance

Quality Assurance is defined by Webster

as “a program for the systematic monitor-

ing and evaluation of the various aspects of a project, service, or facility to ensure

that standards of quality are being met”.

In PERICLES we define a series of Quality Assurance (QA) criteria for the entities

of evolving ecosystems, in particular for policies, processes, complex digital media

objects, semantics and user communities. This will allow us to manage change in the

ecosystem by validating its entities, detecting conflicts and keeping track of its evolu-

tion through time. Our methods focus on validating the correct application of policies

to the ecosystem. When change occurs, the approach will ensure that policies are still

correctly implemented by tracing the correct application of the higher-level policies

(guidelines, principles, constraints) in the concrete ecosystem implementation. Poli-

cies will be expressed at different levels, using a policy model integrated in the Eco-

system Model itself. We support the QA of policies by defining criteria and methods

that can validate or measure the correct application of policies through processes,

services and other ecosystem entities, so ensuring that the implementation is respect-

ing the principles defined in the high-level policies. The QA methods in turn support

the management of change in the ecosystem entities, such as change in policy, policy

https://github.com/pericles-project/PeriCAT

http://www.merriam-webster.com/dictionary/quality%20assurance

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

lifecycle, change in the processes implementing those policies, or change in other

policy dependencies. In this model, we do not make any strong assumption about

the format in which the policy has to be expressed, be it natural or formal language,

nor are we imposing any specific structure on the processes used to implement them

(although we are providing exemplar implementations). We are aware that policies

and processes in real systems will be implemented using a variety of techniques and

we aim to develop a policy layer that can be applied on top of existing ecosystems.

This assumption will allow the deployment of such QA methods in systems that are

not built using only specific technologies or rule languages, making their adoption

simpler.

Other projects have looked at the issue of QA in preservation, in particular the

SCAPE project

, with a focus on the QA of a specific type of digital object (image,

audio, e-publications), and also on the implementation of digital preservation policies

by collecting metrics on collections; this is a valuable approach that is specific to

issues related to digital object QA. Our approach works at the model level and ad-

dresses the implementation of digital preservation policies in existing ecosystems,

although it takes into account the valuable work done by SCAPE.

More concretely, in [8] we define a policy model, a policy derivation process with

guidelines, and a series of QA criteria and change management approaches for poli-

cies, taking into account the possibility of conflicting policies. We are currently work-

ing on exemplar implementation and refinement of the methods, implemented using

the LRM and digital ecosystem models and other PERICLES technologies.

5.3 Appraisal

Appraisal is a process that in broad terms aims to determine which data should be held

by an organisation. This can include both decisions about accepting data for a collec-

tion or archive (e.g. acquisition) as well as determining whether existing data in an

archive or a collection should be retained.

In traditional paper-based archival practice, appraisal is a largely manual process,

which is performed by a skilled archivist or curator. Although archivists are guided by

organisational appraisal policies, such policies are mostly high-level and do not in

themselves provide sufficiently detailed and rigorous criteria that can directly be trans-

lated into a machine executable form. Thus, much of the detailed decision-making

rests with the knowledge and experience of the archivist or curator.

With the increasing volumes of digital content in comparison to analogue, manual

appraisal is becoming increasingly impractical. Thus there is a need for automation

based on clearly defined appraisal criteria. At the same time, decisions about acquisi-

tion and retention are dependent on many complex factors. Hence our aim here is to

identify opportunities for automation or semi-automation of specific criteria that can

assist human appraisal.

In [8], we set out our overall approach to appraisal. In Appendix 1, we extend the

categorisation of appraisal criteria in the DELOS project. The latter focused on ap-

praisal as the determination of the worth of preserving information, that is, as a means

http://wiki.opf-labs.org/display/SP/SCAPE+Policy+Framework

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

of answering the question, ‘what is worth keeping’? Here appraisal is considered as a

process to be revisited throughout the life of the digital object. Consequently, our

results differ principally in the breadth of material considered and in the number and

breadth of appraisal factors identified.

Within the context of the PERICLES case studies, appraisal can naturally be parti-

tioned into two distinct categories.

● Technical Appraisal – decisions based on the (on-going) feasibility of pre-

serving the digital objects. This involves determining whether digital objects

can be maintained in a reusable form and in particular takes into account obso-

lescence of software, formats and policies.

● Content-based (or intellectual) Appraisal – acquisition and retention deci-

sions or assignment of value based on the content of the digital objects them-

selves.

Our focus in this paper is on the technical appraisal aspect. We are primarily inter-

ested here in predictive rather than reactive approaches to modelling the impact of

change. Projects such as PLANETS [29] used a technology watch to detect changes in

the external environment, which could then result in changes to archived content. We

aim to model risks through understanding longer-term trends to predict the impact of

changes in the future. This work follows a number of steps:

● Quantify primary risks to the ecosystem. This is done by analysis and model-

ling of external data sources to predict the likely obsolescence of software

and formats, or hardware failure of entities in the ecosystem.

● Determine the impact of primary risks on entities in the ecosystem. This step

aims to identify entities at the greatest primary risk.

● Determine the impact resulting from higher-order risks propagating through

ecosystem. In this step we propagate risks through the models.

● Determine potential mitigating actions and their associated costs. In some cas-

es, it may be possible to execute and validate mitigating actions automatical-

ly.

The overall goal is to provide a tool for use e.g. by archivists, to analyse a digital

ecosystem, determine at what point in the future there is a significant risk to reuse, and

the potential cost impact and potential mitigating actions. Such a tool could be applied

for example to assess the value of a software-based artwork, by determining how long

it can be displayed in exhibitions before elements become obsolete or require refactor-

ing, or the cost of maintaining a set of scientific experiments for a given time period.

In order to enhance the user experience of the model-driven approach, PERICLES

is developing a visualisation tool MICE (Model Impact and Change Explorer), which

aims to present risk and impact information to users.

6 PERICLES Integration Framework

The PERICLES integration framework, described in detail in [9], is designed for the

flexible execution of varied and varying processing and control components in typical

preservation workflows, while itself being controllable by abstract models of the over-

all preservation system. It is the project’s focal point for connecting tools, models and

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

application use cases together to demonstrate the potential of model-driven digital

preservation.

The integration framework can be deployed in slightly different ways to suit test-

ing and development, or real-world deployment. Fig. 8 shows the configuration for

real-world deployment.

Fig. 8. Real world deployment of PERICLES components, based on the integration framework.

The Entity Registry Model Repository (ERMR) is a component for the management

of digital entities and relationships between them. Access methods are presented as

RESTful services. The ERMR registers and stores entity metadata and models.

Agreed metadata conventions provide the necessary registry functionality; the registry

is agnostic to the entities and metadata stored with the interpretation of data being the

responsibility of the client applications. The ERMR provides a CDMI implementation

for HTTP access to entities in the registry.

The ERMR uses a triple store as a database for the storage and retrieval of triples

through semantic queries. The ERMR also provides a mediator service to integrate

semantic services to extend its reasoning capabilities. It also provides a simple REST-

ful API for access to the registry. Queries can be expressed in the SPARQL query

language to retrieve and manipulate data stored in the triple store. To link entities

described as triples to actual digital objects, the ERMR is able to provide unique iden-

tifiers used to create a unique CDMI URL for an associated object. This URL is used

to link entities stored in data storage components.

The Process Compiler (PC) supports the translation and reconfiguration of preser-

vation process models described in the ERMR into executable workflows to be em-

ployed by the Workflow Engine. As part of this the PC will transfer information to the

ERMR, which is used to update the process descriptions. The current component

implementation is targeted towards a BPMN compilation system, though in theory the

design can be adapted to any workflow engine language.

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

The Workflow Engine takes processes, compiled by the Process Compiler with de-

scriptions and implementations stored in the ERMR, Data Storage and Workflow

Engine cache, and executes them through orchestration of executable components

(PERICLES tools, archive subsystems or supporting software) wrapped in Web Ser-

vices Handlers with REST targets. This is one of the main fixed points of the architec-

ture, which is active at all times.

Processing Elements (PEs) are any pieces of software or hardware that can be used

by a PERICLES process to accomplish a given task. They are characterised by a fixed

set of parameters including input and output types, platform requirements and version-

ing information.

Handlers are at the core of the integration framework. They function as the com-

munication points for each major entity within the system. The Handlers deal with the

validation of incoming requests, exercise the functionality of the PEs they wrap, store

and transfer the results of PE functions and initiate necessary communications with

other PEs. The Handlers do not perform any operations that change or alter the data

contained in the objects they handle; only PEs can alter and change data. PEs, on the

other hand, should not have knowledge of anything in their environment. This means

they perform their function and only their function, and the Handlers deal with the rest

of the PERICLES system and the outside world.

Data Storage is a specialised long-term Processing Element, a permanent service

available to a PERICLES-based system. This component is responsible for storing

digital objects, which can be data files, metadata, models or ontologies. Data Storage

must be represented in the framework as a long-term service, since Processing Ele-

ments are typically transient in nature with limited functionality scope. The Data Stor-

age service must manage data, as required, as bit-level preservation, object replication

and distributed storage mechanisms.

7 Application of PERICLES Technologies to the Wider

Community

As PERICLES is an inter-disciplinary project cutting across a number of technologies

and aimed at diverse communities each with their specific remit, there are several

approaches that the project is adopting for the exploitation of the results [1]. These

include:

 Software products, e.g. component or system level modules.

 Services, e.g. on demand cloud-based preservation services.

 Consulting, e.g. advice on best practices and forming a preservation strategy.

 Training, e.g. commercial training.

 Education courses, e.g. Masters level courses taught at academic institutions.

 Technology licensing, e.g. through the use of patents.

The two application domains, space science and digital media are the primary tar-

gets for technology transfer due to the partner involvement and knowledge.

In space science, Europe (especially ESA) has made large investments over the

years to develop, launch and operate missions in different fields of science (e.g. Earth

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

Observation, Planetary Science, Astronomy, ISS experiments). This has resulted in

long time series and large volumes of data being requested and made available to

different scientific users. However, no explicit mechanisms exist today to cover their

maintenance long after the completion of the operations phase of the relevant missions

and preservation is addressed diversely and only on a mission-per-mission basis.

These data are unique and irreplaceable and constitutes a capital for Europe that is

fundamental to generate economic and scientific advances.

The requirements for libraries, museums, galleries, and archives (and other herit-

age sectors) have evolved quite radically during the last fifteen years. Museums and

galleries are increasingly collecting digital objects. These may be software-based

artworks, design objects or digital objects related to the history of science and engi-

neering. In the case of artworks, these will be acquired and preserved during their

active life and will in the majority of cases evolve and change over time, for example

for display in different operational settings. In other cases there may be a desire to

keep a particular digital object functioning in a way that represents the historical con-

text of that object. In either case, understanding and documenting what those objects

are dependent on and how the digital environments and the objects change over time

is essential to the mission of the museum. The user expectation is that this type of

content will be permanently accessible and valuable. This type of demand is coupled

with the expectation that archived content can be viewed in the “original form”, inde-

pendently of specific software or hardware technologies, thereby re-creating an “au-

thentic” user experience, even if they are associated with software or hardware that is

non-standard or obsolete.

Beyond the space science and cultural heritage sectors, media production is a

growing sector. The business case for preservation in this environment is often based

on the re-use of material in new productions; avoiding the expensive or at times im-

possible task of re-capturing material.

Digital library services provide the infrastructure to underpin teaching and learn-

ing; research and scholarly communication; web services; and other discovery services

based on resource sharing across university and educational sectors. Solutions are

required address the rapid growth and evolution of technology, formats, and dissemi-

nation mechanisms. Preservation requires the tools to provide access, support authen-

ticity and integrity, and address the mitigating effects of technology or media obsoles-

cence.

Projects in Science and Engineering are expensive to setup, or the situation for the

project are unique. Funding agencies are increasingly seeking to ensure the data col-

lected by these projects is kept long enough for any interested groups to make use of

the data. For example, the UK Engineering and Physical Science Research Council

(EPSRC) have started to require that collected data is kept usable for a minimum of 10

years

, with many other funding agencies taking similar approaches. Although such

policies exist, many of the funded research groups lack the expertise or the tools to

meet this requirement.

Increasingly the area of healthcare is adopting ICT to store patient information and

to aid in administering treatment

. By its very nature patient healthcare information

http://www.epsrc.ac.uk/about/standards/researchdata/expectations/

http://www.digitalpreservationeurope.eu/publications/briefs/security_aspects.pdf

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

is highly sensitive and subject to stringent policies that often differ across countries, or

in some cases regions on how the data must be managed (including who can access the

information and how it is disposed) making it very difficult, if not impossible, to dis-

tribute records electronically from one domain to another.

8 Conclusions

The current paper has provided an overview of some of the work being performed by

the PERICLES project, at the end of the third year of the four-year project. The results

we have obtained so far illustrate the potential value of models in digital preservation.

The final year of the project will be focused on producing integrated prototypes and

further developing the user-facing components such as appraisal.

The outcomes of the project align with the EU Digital Agenda for Europe

in sup-

porting the digital preservation of digital cultural assets, and are potentially applicable

across a wide range of sectors beyond the space science and cultural heritage.

References

1. PERICLES Consortium: Deliverable D10.1 – Initial version of exploitation Plan (2014).

http://pericles-project.eu/uploads/files/PERICLES_WP10-D10_1-Exploitation_Plan-V1.pdf

2. Lagos, N., Waddington, S., Vion-Dury, J.-Y.: On The Preservation Of Evolving Digital

Content - The Continuum Approach And Relevant Metadata Models, 9th Metadata And

Semantics Research Conference (MTSR 2015), Manchester, UK. http://pericles-project.eu/

uploads/files/PreservEvolvingDigitalContent-LagosWaddingtonVion-Dury-32-

MTSR2015.pdf

3. Vion-Dury, J-Y, Lagos, N., Kontopoulos, E., Riga, M. Mitzias, P., Meditskos, G., Wad-

dington, S., Laurenson, P., Kompatsiaris, I.: Designing for Inconsistency – The Dependen-

cy-based PERICLES Approach, First International Workshop on Semantic Web for Cultur-

al Heritage (SW4CH 2015), Futuroscope, France. http://www.xrce.xerox.com/

Research-Development/Publications/2015-052

4. PERICLES Consortium: Deliverable D2.3.2 - Data Surveys and Domain Ontologies,

(2015). http://pericles-project.eu/uploads/files/PERICLES_WP2_D2_3_2_Data_Survey_

Domain_Ontologies_V1_0.pdf.

5. Waddington, S., Tonkin, E., Palansuriyam, C., Muller, C., Pandey, P.: Integrating Digital

Preservation into Experimental Workflows for Space Science, PV2015, Darmstadt, Germa-

ny. http://pericles-project.eu/uploads/files/PERICLES_PV2015_KCL_Presentation.pdf.

6. Mitzias, P., Riga, M., Waddington, S., Kontopoulos, E., Meditskos, G., Laurenson, P. and

Kompatsiaris, I.: An Ontology Design Pattern for Digital Video, Proceedings of the 6th

Workshop on Ontology and Semantic Web Patterns (WOP 2015) co-located with the 14th

International Semantic Web Conference (ISWC 2015), Vol. 1461, Bethlehem, Pensylvania,

USA.

http://www.ombudsman.org.uk/__data/assets/pdf_file/0016/24631/Digital-Preservation-

Policy.pdf

https://ec.europa.eu/digital-agenda/en/digitisation-digital-preservation

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies

7. Corubolo, F., Eggers, A., Hasan, A., Hedges, M., Waddington, S, Ludwig, J.: A pragmatic

approach to significant environment information collection to support object reuse, iPres

2014, Melbourne, Australia. http://pericles-project.eu/uploads/files/ipres2014_PET.pdf

8. PERICLES Consortium: Deliverable D5.2 – Basic tools for Digital Ecosystem management

(2015). http://pericles-project.eu/uploads/files/PERICLES_WP5_D5_2_Basic_Tools_for_

Ecosystem_Management_V1_0.pdf.

9. PERICLES Consortium: Deliverable D6.4 – Final version of integration framework and

API implementation (2015). http://pericles-project.eu/uploads/files/PERICLES_WP6_

D64_Final_Version_of_framework_V1_0.pdf.

10. Higgins, S.: The DCC Curation Lifecycle Model. International Journal of Digital Curation.

3, (1), 134–40 (2008). Available from: http://ijdc.net/index.php/ijdc/article/view/69.

11. Ball, A.: Review of Data Management Lifecycle Models (version 1.0). REDm‐ MED Pro-

ject Document redm1rep120110ab10. Bath, UK: University of Bath. (2012). http://opus.

bath.ac.uk/28587/1/redm1rep120110ab10.pdf .

12. Committee on Earth Observation Satellites Working Group on Information systems and

Services (WGISS): Data Life Cycle Models and Concepts CEOS 1.2 (2012).

http://ceos.org/document_management/Working_Groups/WGISS/Interest_Groups/Data_St

ewardship/White_Papers/WGISS_DSIG_Data-Lifecycle-Models-And-Concepts-v13-

1_Apr2012.docx.

13. CCSDS - Consultative Committee for Space Data Systems: Reference Model for an Open

Archival Information System (OAIS), Recommended Practice, CCSDS 650.0-M-2 (Ma-

genta Book) Issue 2 (2012).

14. Upward, F.: Structuring the records continuum (Series of two parts) Part 1: post custodial

principles and properties. Archives and Manuscripts. 24 (2) 268-285 (1996).

http://www.infotech.monash.edu.au/research/groups/rcrg/publications/recordscontinuum-

fupp1.html.

15. McKemmish, S.: Placing records continuum theory and practice. Archival Science Volume

1, Issue 4, 333-359 (2001).

16. PERICLES Consortium: Deliverable D5.1.1 – Initial Report on Digital Ecosystem Man-

agement (2014). http://pericles-project.eu/uploads/files/PERICLES_D5_1_1-Preservation_

Ecosystem_Management_V1_0.pdf

17. Tzitzikas, Y.: Dependency Management for the Preservation of Digital Information. Data-

base and Expert Systems Applications, pp. 582-92. Springer Berlin Heidelberg (2007).

18. Tzitzikas, Y., Marketakis, Y., Antoniou, G.: Task-Based Dependency Management for the

Preservation of Digital Objects Using Rules. Artificial Intelligence: Theories, Models and

Applications, pp. 265–74. Springer Berlin Heidelberg (2010).

19. Marketakis, Y., Tzitzikas, Y.: Dependency Management for Digital Preservation Using

Semantic Web Technologies. Int. Journal on Digital Libraries 10(4): 159-77, (2009).

20. Vorontsov K. V., Potapenko A. A.: Additive Regularization of Topic Models. Machine

Learning. Special Issue “Data Analysis and Intelligent Optimization with Applications”,

101(1), 303-323, (2015).

21. Wittek, P., Darányi, S., Liu, Y.H.: A vector field approach to lexical semantics. In Proceed-

ings of Quantum Interaction-14, Filzbach (2014).

22. Wittek, P., Darányi, S., Kontopoulis, S., Moysiadis, T., Kompatsiaris, I.: Monitoring Term

Drift Based on Semantic Consistency in an Evolving Vector Field. In Proceedings of

IJCNN-15 (2015). http://arxiv.org/abs/1502.01753.

23. Hedges, M. and Blanke, T.: Digital Libraries for Experimental Data: Capturing Process

through Sheer Curation. In Research and Advanced Technology for Digital Libraries (pp.

108-119). Springer Berlin Heidelberg (2013).

24. Geoffrey I. Webb, G.I., Hyde, R., Cao, H., Nguyen, H-L., Petitjean, F.: Characterizing

Concept Drift. Accepted for publication in Data Mining and Knowledge Discovery on De-

cember 10, 2015. At http://arxiv.org/pdf/1511.03816v5.pdf

PERICLES â

A¸S Digital Preservation through Management of Change in Evolving Ecosystems

25. Darányi, S., Wittek, P., Konstantinidis, K., Papadopoulos, S.: A potential surface underly-

ing meaning? Yandex School of Data Analysis Conference, Machine Learning: Prospects

and Applications, Berlin (2015). At https://www.youtube.com/watch?v=PEnRqv9hyzg.

26. Lerman, K., Galstyan, A., Ver Steeg, G., Hogg, T.: Social Mechanics: An Empirically

Grounded Science of Social Media. In Proceedings of International AAAI Conference on

Web and Social Media Fifth International AAAI Conference on Weblogs and Social Media

(2011). http://www.aaai.org/ocs/index.php/ICWSM/ICWSM11/paper/view/3836/4393.

27. PERICLES Consortium: Deliverable D2.3 – Media and Science Case Study Functional

Requirements and User Descriptions (2014). http://www.pericles-project.eu/uploads/files/

PERICLES_D231_Case_studies-Functional_Requirements_gathering.pdf.

28. PERICLES Consortium: Deliverable D4.2 Encapsulation of Environmental Information

(2015). http://www.pericles-project.eu/uploads/PERICLES_WP4_D4_2_Encapsulation_

of_Environmental_Information_V1.pdf.

29. Aitken, B., Helwig, P., Jackson, A., Lindley, A., Nicchiarelli, E., Ross, S.: The Planets

Testbed: Science for Digital Preservation, The Code4Lib Journal, Issue 3, (2008).

http://www.dcc.ac.uk/resources/briefing-papers/technology-watch-papers/planets-

testbed#sthash.mRsLQkcg.dpu.f

EPS Colmar 2015 2015 - The Success of European Projects using New Information and Communication Technologies