=Paper=
{{Paper
|id=Vol-2821/paper8
|storemode=property
|title=A Framework for the Automatic Adaptation of RDF-based Semantic Annotations
|pdfUrl=https://ceur-ws.org/Vol-2821/paper8.pdf
|volume=Vol-2821
|authors=Enio de Jesus Pontes Monteiro,Julio Cesar dos Reis
|dblpUrl=https://dblp.org/rec/conf/semweb/MonteiroR20
}}
==A Framework for the Automatic Adaptation of RDF-based Semantic Annotations==
https://ceur-ws.org/Vol-2821/paper8.pdf
A Framework for the Automatic Adaptation of
RDF-based Semantic Annotations
Enio de Jesus Pontes Monteiro [0000−0001−9992−864X] and Julio Cesar dos
Reis [0000−0002−9545−2098]
Institute of Computing, University of Campinas, Campinas, SP, Brazil
eniojpmonteiro@hotmail.com, jreis@ic.unicamp.br
Abstract. Access and use of semantically defined metadata based on RDF repos-
itories can benefit several types of computational tasks. However, RDF triples
tend to undergo modifications as new releases of the repositories appear, which
implies a challenging scenario for RDF-based generated annotations. In this con-
text, existing annotations need to be updated according to the evolution of un-
dergoing knowledge base used for their definitions. In this paper, we propose
an adaptation framework for updating semantic annotations defined from struc-
tured RDF data. Our adaptation approach relies on modifications detected in the
evolution of RDF knowledge bases. We design and formalize adaptation opera-
tions which are applied to update annotation states. We present and formalize the
framework and discuss existing open challenges in our research task.
Keywords: Metadata · RDF · Ontology · Semantic Annotations · LOD
1 Introduction
The generation of metadata (data about data) on Web documents, videos and images
using existing Resource Description Framework (RDF) knowledge bases plays a key
role to computational systems. This type of metadata is called semantic annotations [1]
and it consists of RDF resources that make the meaning of Web elements interpretable
to machines. Semantic annotations are essential elements to help systems better in-
terpret, integrate, and retrieve information considering the explicit meaning shared by
machines.
In the last years, a large number of interconnected RDF knowledge bases have
emerged describing various types of resources in a structured way, e.g., Dbpedia [2].
The knowledge presented in knowledge bases described in RDF is constantly evolving.
This phenomenon of the evolution of the base can directly affect the existing associated
annotations (already created) since it can make them invalid.
Existing literature has presented methodologies to address this problem. The stud-
ies that address the automatic detection of inconsistent annotations [3–6] perform the
identification of concepts that changed from a release j of a Knowledge Organization
Systems (KOS) [7] to its release j + 1 and its associated set of annotations. These
studies do not support the correction of outdated annotations. However, methods have
Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
been developed to address the maintenance of outdated annotations [8–11]. Neverthe-
less, our literature analysis did not detect investigations addressing the maintenance of
annotations considering their generation at the instance level of concepts.
In this article, we propose a framework capable of identifying and applying main-
tenance actions in semantic annotations affected by the evolution of RDF knowledge
bases as automatically as possible. Our maintenance process comprises the execution
of three steps (cf. Secion 2).
The remaining of this article is organized as follows: Section 2 presents our frame-
work including its description and formalization; Section 3 provides a discussion on
existing open challenges in our research. Finally, Section 4 draws the conclusion re-
marks.
2 ANNOLOD framework for annotation adaptation
The key contribution of this research consists of a framework capable of executing
modifications to update annotations as automatically as possible. We propose the frame-
work ANNOLOD to support maintaining instance annotations in RDF repositories. In
our study, we adapted the annotation model proposed by Cardoso et al. [6] in order
to consider instance-based generated annotations. We defined our model as ISAM =
(D, Oj , Rj , A, SemRel, Uf ), such that:
– D: It consists of a set of documents D = {dj, ..., dn}.
– Oj : is a ontology in its release j. An ontology O describes a domain of knowl-
edge in terms of concepts, attributes, and relationships between concepts [13]. For-
mally, an ontology O = (CO , SO , PO ) consists of a set of classes CO interrelated
by directed relationships SO . Each c ∈ CO concept has a unique identifier and is
associated with a set of attributes PO (c) = {a1 , a2 , a3 , ..., an }.
– Rj : is a RDF repository in its release j with its predicates defined in the ontol-
ogy Oj . An RDF repository in the context of Linked Open Data (LOD) is a finite
set of RDF triples [12]. Formally, R = (t1 , t2 , t3 , ..., tn ). In a RDF repository, a
triple associates two nodes (or resources) using a property (predicate). A resource
can be defined as an instance of a certain ontology class. In RDF, resources are
described using a Uniform Resource Identifier (URI)1 for the unique identification
of resources on the Web. A RDF triple refers to a data entity composed by subject
(s), predicate (p), and object (o) defined in the form of t = (s, p, o).
– A: is a set of annotations. A a ∈ A is defined as a = (i, t, d, Offset, rel, p), such
that, an entity named i ∈ d ⊂ D is connected to a triple t ∈ Rj ; Offset indicates
the position (start, end) where i appears in the document d being annotated; rel ∈
SemRel describes the type of relationship between i and s ∈ t; p ∈ Uf points out
to the previous version of the annotation ai to keep a tracing of the evolution of the
annotation in time.
The adaptation was carried out, because the model defined in Cardoso et al. [6] did
not have all the required elements to conceive our maintenance actions. We observed
1
https://www.w3.org/wiki/URI
the need to add the attributes i, t, d, Of f set, rel, and p in the definition of an annotation
a. In the original model of Cardoso et al. [6], they were defined at set level of A. Table
1 provides an instance annotation example by adopting our adapted model (ISAM).
The mention of the scientist “Albert Einstein”, present in a given textual document was
linked (annotated) to its semantic definition (formal RDF resource “Albert Einstein”)2
formally coded in the DBpedia.
Table 1. Annotation example based on our ISAM model
A D Rj → O j SemRel Uf
of f set
i d t rel p
a3 start end
Albert Einstein 46 1 15 sameAs a2
The proposed framework defines annotation adaptation actions to be performed au-
tomatically when an RDF dataset used to create semantic annotations evolves (i.e., a
new release is generated). These actions are necessary to keep annotations consistent
and up to date over time. The necessary input consists of the interconnected initial
datasets, being Rj and Rj+1 (its new release that can affect existing annotations) and
the existing in place Aj annotations. The ultimate goal of the framework is to obtain the
updated Aj+1 annotations according to the new release Rj+1 dataset. The framework
performs a series of steps (cf. Algorithm 1). Each step is explained in further details (cf.
Steps A, B, and C).
Algorithm 1: Annotation Maintenance
Require: Rj , Rj+1 , Aj
1: Aj+1 ← ∅
2: ∆ ← detectChanges(Rj , Rj+1 )
3: Aaf f ← recAf f Annotations(∆, Aj )
4: Aunaf f ← recU nAf f Annotations(∆, Aj )
5: Aj+1 ← Aj+1 ∪ Aunaf f
6: for all ai ∈ Aaf f do
7: Aj+1 ← Aj+1 ∪ applyAction(∆, ai )
8: end for
9: return Aj+1
Step A: this step consists of detecting a series of modifications that occurred in a
given time period based on two releases of an RDF dataset (line 2 in Algorithm 1). This
operation is known as ∆ because it computes the difference between the two datasets,
recognizing added, removed, and not updated elements. Changes can be of the simple
type (such as unit actions of adding or removing triples), or complex operations (update
actions) of the knowledge stored in the datasets.
Step B: this step consists of recognizing and filtering the annotations affected by
the changes of those that are not affected (lines 3 to 4 in Algorithm 1). An annotation
2
http://dbpedia.org/page/Albert_Einstein
can be created, removed or updated. In our solution, annotations that share a subject
(s) with a tk triple in which tk ∈ ∆ are considered affected by change modifications.
These are considered outdated annotations and maintenance actions on the annotations
must be applied to them. We assume that unaffected annotations can be directly reused
in composing the final set of updated annotations and added to the final set of Aj+1
annotations (line 5 in Algorithm 1). However, annotations classified as affected is fur-
ther handled by our framework. This involves investigating which and how computed
change operations influence the definition of existing annotations.
Step C: this step involves applying corrective actions (lines 6 to 8 in Algorithm 1)
to the affected and outdated annotations detected in step B. For example, an action type
may be a “reanotation”. In this case, an annotation ai ∈ Aj defined on the basis of a
triple tk adapted its subject (s). The framework generates as a final result a stable and
semantically consistent set of annotations Aj+1 (line 9 in Algorithm 1) concerning the
updated RDF data in the new dataset Rj+1 .
3 Open Research Challenges
The key research challenge at step A (cf. Section 2) is to understand what kind of
changes at the level of instances can affect existing annotations. For example, to what
extent does the removal of a triple RDF (used in defining an annotation) impact the
consistency of such an annotation?
In step B (cf. Section 2), a key challenging research refers to how to accurately
categorize an annotation as inconsistent based on computed change operations. In this
sense, we need to investigate to which extent semantics defined in the annotations are
affected by the observed changes. For example, the removal of the associated triple
may be a typical case that affects the annotation. However, if there are other types of
changes related to the triple subject, it is necessary to further understand how they make
the annotation semantically inconsistent.
The main challenge in step C (cf. Section 2) refers to the definition and correct
application of annotation adaptation actions to ensure the updating of semantically con-
sistent affected annotations. The definition of actions (adaptation operations) requires
investigating techniques that express the necessary conditions for their application.
4 Conclusion
The real value of semantically-enabled computer systems lays on the reliability of se-
mantic annotations. This work studied how to keep RDF-based annotations up-to-date
according to the evolution of RDF repositories. We proposed a framework for the (semi-
)automatic maintenance of semantic annotations affected by RDF data evolution. Our
defined adaptation algorithm works on the basis of change operations automatically
identified in the evolution of RDF datasets. We are currently further investigating the
adaptation actions, their formalization and applicability. Next steps involve the full im-
plementation of a software tool for the adaptation of RDF-based semantic annotations
maintenance. We also plan to conduct thorough experimental analyses with real-world
datasets.
Acknowledgments
This work is supported by the São Paulo Research Foundation (FAPESP) (Grants #2017/02325-
5, #2019/14582-8 and #2013/08293-7)3 .
References
1. Oren, E., Möller, K., Scerri, S., Handschuh, S., Sintek, M.: What are semantic annotations?.
DERI Galway, 9, 62 (2006)
2. Auer S., Bizer C., Kobilarov G., Lehmann J., Cyganiak R., Ives Z.: DBpedia: A Nu-
cleus for a Web of Open Data. In: Aberer K. et al. (eds) The Semantic Web. ISWC
2007, ASWC 2007. LNCS, vol 4825, pp 722-735. Springer, Berlin, Heidelberg (2007).
https://doi.org/10.1007/978-3-540-76298-0_52
3. Maynard, D. Peters, W., d’Aquin, M., Sabou, M.: Change management for metadata evolution.
In: Proceedings of the International Workshop on Ontology Dynamics (IWOD-07), pp. 27–40.
IWOD-07, Innsbruck (2007)
4. Gross A., Hartung M., Kirsten T., Rahm E.: Estimating the Quality of Ontology-Based An-
notations by Considering Evolutionary Changes. In: Paton N.W., Missier P., Hedeler C. (eds)
Data Integration in the Life Sciences. DILS 2009. LNCS, vol 5647, pp 71-87. Springer, Berlin,
Heidelberg (2009). https://doi.org/10.1007/978-3-642-02879-3_7
5. Köpke J., Eder J.: Semantic Invalidation of Annotations Due to Ontology Evolution.
In: Meersman R. et al. (eds) On the Move to Meaningful Internet Systems: OTM
2011. OTM 2011. LNCS, vol 7045, pp 763-780. Springer, Berlin, Heidelberg (2011).
https://doi.org/10.1007/978-3-642-25106-1_25
6. Cardoso, S. D., Pruski, C., Da Silveira, M., Lin, Y. C., Groß, A., Rahm, E., Reynaud-
Delaître, C.: Leveraging the Impact of Ontology Evolution on Semantic Annotations. In:
Blomqvist E., Ciancarini P., Poggi F., Vitali F. (eds) Knowledge Engineering and Knowl-
edge Management. EKAW 2016. LNCS, vol 10024, pp 68-82. Springer, Cham (2016).
https://doi.org/10.1007/978-3-319-49004-5_5
7. Hodge, G.: Systems of Knowledge Organization for Digital Libraries: Beyond Traditional
Authority Files. 1st edn. ERIC, Washington, DC (2000)
8. Luong PH., Dieng-Kuntz R. (2006) A Rule-based Approach for Semantic Annotation Evo-
lution in the CoSWEM System. In: Koné M.T., Lemire D. (eds) Canadian Semantic Web.
Semantic Web and Beyond (Computing for Human Experience), vol 2, pp 103-120. Springer,
Boston, MA . https://doi.org/10.1007/978-0-387-34347-1_7
9. Park, Y.R., Kim, J., Lee, H.W. et al.: GOChase-II: correcting semantic inconsistencies from
Gene Ontology-based annotations for gene products. BMC Bioinformatics 12, S40 (2011).
10. Cardoso, S. D., Chantal, R. D., Da Silveira, M., Pruski, C.: Combining rules, background
knowledge and change patterns to maintain semantic annotations. In: AMIA Annual Sympo-
sium Proceedings, pp. 505–514. AMIA, Washington, D.C (2017)
11. Cardoso, S., Reynaud-Delaître, C., Da Silveira, M., Lin, Y. C., Gross, A., Rahm, E., Pruski,
C.: Evolving semantic annotations through multiple versions of controlled medical terminolo-
gies. Health Technol 8, 361–376 (2018)
12. Faisal S., Endris K.M., Shekarpour S., Auer S., Vidal ME.: Co-evolution of RDF Datasets.
In: Bozzon A., Cudre-Maroux P., Pautasso C. (eds) Web Engineering. ICWE 2016. LNCS, vol
9671, pp 225-243. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-38791-8_13
13. Gruber, T. R.: Toward principles for the design of ontologies used for knowledge sharing?.
International journal of human-computer studies, 43(5-6), 907-928 (1995)
3
The opinions expressed in this work do not necessarily reflect those of the funding agencies.