Semantic Preventive Conservation of
Cultural Heritage Collections
Efthymia Moraitou1, John Aliprantis1 and George Caridakis1
1 University of the Aegean
School of Social Sciences, Department of Cultural Technology and Communication,
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org
Abstract. Semantic knowledge has been proven to be rather efficient on the da-
ta management of Culture Heritage domain. Conservation is an important as-
pect of museum management cycle aiming to preserve cultural heritage objects
in the best possible condition for future generations. Since cultural objects are
susceptible to environmental changes, sensor data could be of significant impor-
tance in automatic environmental monitoring and possible conservation issues.
Recently, many approaches have included the SSN (Semantic Sensor Network)
ontology in their domain knowledge representation and relevant applications.
In this work, we merge the SSN using the CORE (Conservation Reasoning) on-
tology, an ontology which is based on empirical analysis, scientific knowledge
and existing vocabularies of the conservation domain. Incorporating many of
the existing properties of both ontologies and proposing additional ones, we in-
tegrate the majority of SSN classes in the CORE ontology, creating a new
merged ontology that combines conservation procedures data and rules with
sensor and environmental information. Furthermore, we create ontology-based
rules, using the SWRL (Semantic Web Rule Language), in order to express
preventive conservation guidelines and rules based on sensor and object current
Keywords: Conservation Reasoning, SSN Ontology, Ontology Integration,
CIDOC CRM Development, Cultural Heritage.
Artworks conservation is an important process of museum collection management
cycle, aiming to preserve it in the best possible condition for present and future gener-
ations . Conservation procedures1, such as examination, analysis, diagnosis, pre-
ventive or active conservation, require the consistent documentation of diverse infor-
mation which the museum must organize, manage and potentially share. Furthermore,
conservators and scientists of the conservation domain must be aware of related in-
formation in order to reach conclusions and take decisions relevant to their work .
1 Conservation embraces preventive conservation, remedial conservation and restoration.
While the major part of information is generated by the scientists of the domain of
Cultural Heritage (CH) conservation, as a result of their observations and activities,
there are also valuable data which are produced by sensors or sensor networks in the
context of preventive conservation activities. Preventive conservation2 includes indi-
rect actions taken to avoid and minimize future deterioration of artworks and collec-
tions and therefore is related to the management of environmental conditions [1, 3].
Nowadays, metadata standards and schemes, as well as the mapping between them,
facilitate the structural and syntactic interoperability and therefore the information
organization, search and retrieval of the CH conservation domain . An approach of
ontology-based knowledge representation could create a context of intelligent infor-
mation management, defining concepts and their relations, as well as their use in the
semantic web .
Taking into consideration the above mentioned statements, semantic knowledge
and ontology-based rules could be efficient for the management of conservation in-
formation and sensor data. Besides the creation of semantic and interoperable data,
the conceptual representation of domain knowledge could support the ontology-based
rules generation . The ontology-based rules could express preventive conservation
guidelines and rules, combining information related to objects’ condition state, pro-
duction materials and techniques, environmental conditions, damage mechanisms,
causes and results.
Therefore, the use of ontology for artworks conservation and environmental moni-
toring, as well as the expression of ontology-based domain rules which reflect the
knowledge of the discipline is considered beneficial. In the remainder of this paper
we first describe the information management requirements pertaining to artworks
conservation and monitoring, while we provide an overview of related work in the
area of conservation knowledge and semantic sensor data management (Section 2).
Thereafter we describe the integration between CORE (Conservation Reasoning) and
SSN ontology, as well as the expression of ontology-based rules (Section 3). Finally,
we conclude with a brief discussion of future trends regarding to the application of
ontologies and ontology-based rules in the domain (Section 4).
2 Motivation and Related Work
2.1 Documentation and Environmental Monitoring in the Conservation
Artworks and collections present features inseparable to their creation, use and histo-
ry, which are neither always known nor stable. All the original features and changes
must be examined and documented by the scientists of the conservation domain. De-
tailed and accurate documentation in textual (reports) or visual (photographs, dia-
grams, designs etc.) records is necessary . However, different conservation activi-
ties require different ways (analytical or brief) and modes of information recording.
Generally, the information material that scientists of conservation domain collect and
produce may refer to an object condition –before or after conservation treatment– and
2 Preventive conservation includes activities about the storage, handling, exhibition, packing and transpor-
tation, the security and emergency planning.
pathology, production materials and techniques, applied conservation materials and
methods, analysis methods, as well as some administrative information .
In the context of preventive conservation, scientists often use sensors to monitor
and control some critical physical parameters related to the degradation of the art-
works. Sensors are small sensing devices which change their status according to phys-
ical stimulus and can be attached to larger objects or specific location [9, 10]. Never-
theless it must be mentioned that the ideal atmosphere may differ for each item, or
group of items, according to their original materials and current condition state [3,
11]. According to the different cases and requirements, data loggers, as well as sen-
sors in wired or wireless sensor networks, have been used in different implementa-
tions. The provided measurements can be downloaded to a computer and analyzed
from time to time, or in the case of networks communication, data flow from low-
powered devices to high-powered systems (also called platforms) for further aggrega-
tion and processing . Additionally in potential ΙοΤ (Internet of Things) architec-
tures for museums, sensor data are transferred to a cloud by means of gateways .
Commonly, the main environmental factors which are monitored are related to the
temperature and relative humidity, but in some cases sensors are used for the detec-
tion of light and other forms of radiation, the pollution, the pests and the vibration
The combination of documented information and recorded sensor data could im-
prove the conservation specialists’ work. In the short-term, the aforementioned sus-
tained supervision aims to the immediate detection of environmental changes. There-
fore, by using existing knowledge about an object feature it is easy to estimate wheth-
er the condition is harmful, act accordingly and reduce the potential risk [10, 11].
Furthermore the long-term records of sensed data, in relation to other information
which have been documented during conservation procedures, may lead to useful
inferences about the relation between the material decay and its environment .
2.2 Conservation Information and Sensor Data Management
Considering the amount and diversity of information related to conservation proce-
dures, high organization in a concept level is often required for its integration and
management. Conceptual Reference Model (CIDOC CRM) is a widely used ontology
for CH and conservation domain, though not always effective. It has been noticed that
the known information in a particular point of time during conservation documenta-
tion, sometimes cannot be expressed by a CIDOC CRM entity [13, 14].
Conservation is an interdisciplinary science, consequently it is useful to include da-
ta models and ontologies of related domains, such as chemistry domain. In this con-
text OreChem data model and CRMsci (Science Observation Model) have been very
useful for analysis and examination [15, 16]. Nevertheless there are some domain
ontologies about conservation science and procedures. The Ontology of Paintings and
Preservation of Art (OPPRA) draws existing ontologies such as CIDOC CRM, Ore-
Chem and OIA-ORE and aims at the description of chemical analysis/characterization
data . Furthermore, PARCOURS is a domain ontology dedicated to conservation
and restoration domain . Finally, CORE ontology extends CIDOC CRM with
concepts and relations about materials and techniques, condition state and conserva-
tion processes of artworks, and particularly byzantine icons .
Similarly to conservation information, sensor data may be difficult to be shared, in-
tegrated and processed, in order to support knowledge extracting and reasoning capa-
bilities such as intelligent decision making. This is caused because sensor networks
are consisted of devices with increased heterogeneity which produce various types of
data and measurements. To overcome the lack of semantics in sensor networks, se-
mantic technologies are used to automatically annotate and enrich sensor data, add
semantic metadata and information and resolve the heterogeneous of sensor data .
In this direction Semantic Sensor Web (SSW) uses declarative descriptions of sen-
sors, networks and domain concepts to search, query and manage the network and
sensor data .
Sensor ontology is one of the most important components of the SSW . In the
past years there have been developed general sensor ontologies, as well as ontologies
for more specific applications (such as CSIRO, CESN, OntoSensor etc). However,
there were problems in terms of the sensor ontology structure and the expression of
processes and systems’ composition. Therefore, the W3C Semantic Sensor Network
(SSN) Incubator Group proposed a more generic, field-independent model, the SSN
ontology. Developed from developers of the CSIRO, MMI and OOTethys ontologies,
the SSN addresses many of the problems in the older ontologies. The SSN ontology
integrates and upgrades the original ontologies with more detailed classification and a
wider range of generality [20, 21].
As proposed in previous works the semantic sensor data can be connected with
domain concepts related to a specific scenario where the sensor networks are used .
This type of organization may be interesting in order to further analyze data and veri-
fy its compliance with domain rules. A very similar idea has been proposed in WISE-
MUSEUM project specifying art conservation rules .
3 Knowledge Semantic Representation
3.1 Methodology and Tools
The domain ontology for the representation of conservation domain knowledge and
sensor networks concepts was achieved with the integration of CORE and SSN ontol-
ogy. The CORE ontology builds upon and extends the CIDOC CRM ontology, while
is based on empirical analysis, scientific knowledge and existing vocabularies of the
conservation domain. The CORE ontology consists of a base of 11 classes, each of
which branch into subclasses with semantic consistency. CIDOC CRM top-level
classes capture the provenance information about an artwork while the CORE exten-
sions capture the domain related knowledge . On the other hand SSN ontology
consists of 41 concepts and 39 object properties, and can describe sensors, the accura-
cy and capabilities of such sensors, observations and methods used for sensing .
Furthermore, it is built around a central Ontology Design Pattern (ODP) describing
the relationships between sensors, stimulus, and observations, the Stimulus-Sensor-
Observation (SSO) pattern .
CORE development and its integration with SSN entities are achieved with the free
open source software Protégé (Protégé Desktop version 5.2.0) of Stanford University
. Entities attributes and inference rules were also included to support a finer level
of granularity for the domain. In some point the efficiency and consistency of the
ontology was tested by the reasoners “Pellet” and “Hermit”. In addition, rules which
express knowledge of the domain of preventive conservation were formulated in
SWRL rule language and were expressed through SWRL tab of Protégé.
3.2 CORE-SSN Entities and Relations
Considering the amount and diversity of information related to conservation proce-
dures, high organization in a concept level is often required for its integration and
management. In order to model conservation information and sensor data which de-
rived from monitoring of environmental conditions, CIDOC CRM, CORE domain
ontology and SSN sensors ontology were combined. CIDOC CRM is the top-
ontology which CORE entity extends, while the SSN entities are manually mapped
and integrated with the rest of the other two ontologies.
CORE ontology is based on CIDOC CRM classes and furthermore includes enti-
ties which aim to model more accurately the knowledge relating to artworks and col-
lections (a) physical and material structure, (b) pathology, (c) conservation proce-
dures, (d) environment and (e) information resources. Therefore, there are CORE
concepts which organize and represent types of measurement activities related to the
processes of monitoring and material analysis or modification activities related to
processes of sampling and conservation. Furthermore, there are concepts about prop-
erties such as temperature, physical features such as types of material or structure
attributes and deterioration, information objects and so on. An interesting aspect of
CORE concepts structure is the fact that the entities about a damage cause, mechan-
ism and result are separately defined. Therefore, it is possible to capture the informa-
tion about what is observed and what is concluded or could potentially appear as a
consequence. Additionally, the aforementioned possible conclusions or consequences
in some cases were captured as axioms of the ontology.
Some of these concepts are semantically related to the classes of SSN ontology. As
a result it was possible to integrate SSN classes in CORE structure either as sub-
classes, for example SSN class ‘Stimulus’ is the subclass of ‘E5 Event’. Moreover
some SSN classes were defined as equivalent to some CIDOC CRM or CORE
classes, for example SSN class ‘Observation’ was equivalent to CORE class ‘Moni-
toring’, maintaining the semantic consistency. In Figure 1 the main integration and
correlation between SSN and CORE entities is presented.
Fig. 1. Main classes and subclasses of CORE-SSN integration.
The logical association between the CORE and SSN classes was further achieved
by using the existing relations of the ontologies, as well as by adding some new. For
example, an object property was created in order to correlate the SSN class ‘Feature
of Interest’ with a CORE class which corresponds to a concept of object or place of
observation, such as ‘Site’ or ‘Work of Art’. Therefore, the object property ‘equals to’
and its inverse object property ‘is equaled to’ was added (Fig. 2).
In order to test the scope and integrity of CORE-SSN ontology a number of indi-
viduals of entities and object property assertions between them was created. There-
fore, some expressions related to the temperature monitoring of an exhibition hall
were captured. Particularly using the SSN classes and relations, alongside these of
CORE, we could define that the Exhibition Hall was observed, the property of obser-
vation was specifically the temperature, the particular environmental factor of Exhibi-
tion Hall was heat and that the stimulus by which the observation was originated was
triggered by heat change (Fig. 2). The above mentioned information tends to be more
expressive since it captures technical information about sensors function and scientif-
ic information about environmental conditions. Moreover, these concepts and rela-
tions could express the potential mechanisms which could be triggered and the dam-
ages which could be caused, since the factors and phenomena concepts are included
in CORE conservation science ontology.
Fig. 2. A conceptual mapping between entities and relations of CORE-SSN integration.
3.3 Reasoning and Ontology-Based Rules
The aforementioned semantic organization of concepts and relations was used for the
definition of rules and therefore the generation of inferring information. Initially, we
formulated rules in order to further define the relations between concepts. For exam-
ple, in cases of temperature monitoring of a site, in particular an exhibition hall, it is
possible to use both the concepts of heat and temperature. However the first is re-
ferred to the environmental factor while the second to its measured dimension. There-
fore, having defined some basic relations between individuals and formulating the
rule S1, we could have the inferring information that the temperature actually refers to
the heat of the place and that it was observed particularly by a temperature measure-
ment activity. Using the SWRL syntax, the above mentioned rule can be expressed as
shown in Table 1.
Table 1. SWRL Rule 1 (S1).
Core:Heat (?h), core:Exhibition_Hall(?eh), sosa:Observation(?o), core:measured(?o,
?eh), core:Temperature(?t), core:has_environment(?eh,?h),
core:has_dimension(?eh,?t) -> core:has_dimension(?h,?t),
Activating the reasoner Pellet some useful information is inferred, according to rules
and relations. Particularly, the individual Observation1 whose type is the SSN entity
Observation, is inferred that can be equally be defined by the type Tempera-
ture_Measurement which is a CORE entity. The aforementioned inference is due to
the fact that Observation1 is correlated to other individuals which verify the rule S1.
Therefore, SSN and CORE concepts can be equally used for the definition and query-
ing of relevant individuals.
Moreover, the ontology-based rules could express preventive conservation guide-
lines and rules, such as the definition of a temperature “set point” for a sensor of the
system. However, we have to take into account that in practice the set point for an
environmental factor may differ according to the needs and the general condition of
an item, group of items, site etc. For example, the below SWRL expression (S2) uses
the built-in atom “swrlb:greaterThan” to compare the temperature measurement of an
observation with a threshold (ex. 35 °C) and infer that there is a change in heat factor
Table 2. SWRL Rule 2 (S2).
sosa:Observation(?o) , sosa:Result(?r), core:Heat(?h) , core:Heat_Change(?hc) ,
sosa:has_result(?o, ?r), core:has_temperature(?r, ?num) ,
-> core:has_environmental_change(?h, ?hc)
In the context of CORE ontology some axioms about the mechanisms and the results
of environmental changes had been formulated. For example, axioms have expressed
the fact that heat change triggers the physicochemical mechanism of heating and that
heating effects mechanical damage, such as swelling. Furthermore, the expressivity of
the ontology allows the definition of the artworks which may be exhibited in the place
under observation. The structure and production materials of the object can be ex-
pressed with CORE entities and relations. Using CORE ontology classes and rela-
tions, as well as the included axioms, we could formulate rules about the potential
impact of the heat change on the materials and structural layers of an artwork. For
example, we could correlate the potential damage of swelling, which is effected by
heating, with an artwork, which has a textile support layer.
In this case, we used the CORE relation “has the tendency to” in order to express
a rule about the potential presentation of swelling on an artwork with textile support
(Table 3). The information that had been inferred, using the axioms and rules, about
the environmental change of the Exhibition Hall, the possibility of a damage mechan-
ism activation and the impact of this change on an artwork that is exposed in this
condition, could be useful in the context of querying in order to support recommenda-
tions or decision-making.
Table 3. SWRL Rule 3 (S3).
core:Work_of_Art(?w), core:places(?eh,?w), core:Textile_Support(?ts),
core:Exhibition_Hall(?eh), core:has_structural_layer(?w, ?ts), core:Swelling(?sw)
Taking into consideration additional information about sensor measurements, obser-
vations and samples, we can create a system which provides predictions about the
risks and deterioration of the objects regarding environmental conditions such as heat
4 Conclusion and Future Work
In this work, the CORE, an ontology for conservation domain, is integrated with the
SSN ontology in order to combine information about the artworks’ condition state and
environmental conditions, and express preventive conservation guidelines and rules.
In our approach, we integrate the majority of the SSN classes into the CORE structure
either as subclasses, or as equivalent to CORE classes, while also we use the existing
object properties of both ontologies and add a few new, to achieve the logical associ-
ation between them.
The CORE - SSN integration is developed in the open source software Protégé,
and the reasoners “Hermit” and “Pellet” are used for rules implementation. Neverthe-
less, further work is necessary for the validation of the integration and the testing of
the rules efficiency. Furthermore, in regards of the rule language, the SWRL syntax
was used at this stage, though other rule languages are considered to be tested as well
in the future, such as Jena Rule. It is probable that the requirements of real-time
processing, ontology’s complexity and the amount of the processed semantic data
could potentially lead to the use of a different rule language .
Future research could mainly focus on the design and development of a recom-
mendation system that will provide users useful advices and suggestions based on the
semantic rules and information derived from sensor data and objects’ features. By
incorporating systems like Wireless Sensor Networks (WSN) and context-aware ser-
vices and using the CORE-SSN ontology approach, we aim in designing a conserva-
tion system that automatically control environmental conditions according to sensor
data and support decision-making in compliance with art conservation rules and se-
1. International Council of Museums – Committee of Conservation (ICOM - CC), “Termi-
nology to characterize the conservation of tangible Cultural Heritage”. [Online]. Available:
2. Marty PF, Jones KB (2012), Museum Informatics - People, Information, and Technology
in Museums. Routledge, New York, London
3. Perles, A., Pérez-Marín, E., Mercado, R., Segrelles, J. D., Blanquer, I., Zarzo, M., & Gar-
cia-Diego, F. J. (2018). An energy-efficient internet of things (IoT) architecture for pre-
ventive conservation of cultural heritage. Future Generation Computer Systems, 81, 566-
4. Chan LM, Zeng ML (2006) Metadata Interoperability and Standardization - A Study of
Methodology, Part I: Achieving Interoperability at the Schema Level. D-Lib Magazine
5. Gómez-Pérez A, Fernández-López M, Corcho O (2005) Ontological Engineering With Ex-
amples from the Areas of Knowledge Management, e-Commerce and the Semantic Web,
1st edn. Springer-Verlang, London
6. Rattanasawad T, Saikaew K. R., Buranarach M, Supnithi T. (2013) "A review and com-
parison of rule languages and rule-based inference engines for the Semantic Web," 2013
International Computer Science and Engineering Conference (ICSEC), Nakorn Pathom,
2013, pp. 1-6.
7. Letellier R, Schmid W, LeBlanc F (2007) Recording, Documentation, and Information
Management for the Conservation of Heritage Places: Guiding Principles. Los Angeles,
CA: Getty Conservation Institute. http://hdl.handle.net/10020/gci_pubs/recordim. Ac-
cessed 20 Feb 2017
8. Radin M, Živković V (2013) Information Management for Cultural Heritage Conservation.
Преглед НЦД 23:22-32
9. Moraru, A., & Mladenić, D. (2012). A framework for semantic enrichment of sensor data.
Journal of computing and information technology, 20(3), 167-173.
10. Peralta, L. R., de Brito, L. P. L., Gouveia, B. T., de Sousa, D. G., & Alves, C. D. S. (2010).
Automatic monitoring and control of museums environment based on wireless sensor net-
works. Electronic Journal of Structural Engineering.
11. Valach, J., Julis, K., Stefcová, P., Pech, M., Wolf, B., Kotyk, M., & Frankl, J. (2015). Eve-
rything is Data-Overview of Modular System of Sensors for Museum Environment. The
International Archives of Photogrammetry, Remote Sensing and Spatial Information
Sciences, 40(5), 439.
12. Compton, M., Henson, C., Lefort, L., Neuhaus, H., & Sheth, A. (2009, October). A survey
of the semantic specification of sensors. In Proceedings of the 2nd International Confe-
rence on Semantic Sensor Networks-Volume 522 (pp. 17-32). CEUR-WS. org.
13. Le Boeuf, P., Doerr, M., Ore, C. E., & Stead, S. (2015). Definition of the CIDOC concep-
tual reference model. ICOM/CIDOC Documentation Standards Group and CIDOC CRM
Special Interest Group, version, 6(1).
14. Vassilakaki, E., Zervos, S., & Giannakopoulos, G. (2015, February). CIDOC-CRM exten-
sions for conservation processes: A methodological approach. In AIP Conference Proceed-
ings (Vol. 1644, No. 1, pp. 185-192). AIP.
15. Open Archives, “oreChem-Cyberinfrastructure for Chemistry.” [Online]. Available:
16. Doerr, M., Kritsotaki, A., Rousakis, Y., Hiebel, G., & Theodoridou, M. (2014). CRMsci:
the Scientific Observation Model.
17. Hunter, J., & Odat, S. (2011, December). Building a semantic knowledge-base for painting
conservators. In E-Science (e-Science), 2011 IEEE 7th International Conference on (pp.
18. Niang, C., Marinica, C., Markhoff, B., Leboucher, E., Malavergne, O., Bouiller, L., Dar-
rieumerlou, C., & Laissus, F. (2017). Supporting Semantic Interoperability in Conserva-
tion-Restoration Domain: The PARCOURS Project. Journal on Computing and Cultural
Heritage (JOCCH), 10(3), 16
19. Moraitou, E., & Kavakli, E. (2018). Knowledge Management Using Ontology on the Do-
main of Artworks Conservation. Revised Selected Papers. Vol. 10605. Springer.
20. Wang, X., Zhang, X., & Li, M. (2015). A survey on semantic sensor web: sensor ontology,
mapping and query. International Journal of u-and e-Service, Science and Technology,
21. Compton, M., Barnaghi, P., Bermudez, L., GarcíA-Castro, R., Corcho, O., Cox, S., &
Huang, V. (2012). The SSN ontology of the W3C semantic sensor network incubator
group. Web semantics: science, services and agents on the World Wide Web, 17, 25-32.
22. Protégé 2000 Wiki. http://protegewiki.stanford.edu/wiki/Main_Page. Accessed 20 Feb
23. Dell’Aglio, D., Della Valle, E., van Harmelen, F., & Bernstein, A. (2017). Stream reason-
ing: A survey and outlook. Data Science, (Preprint), 1-25.