Semantic Resource Adaptation Based on Generic Ontology Models

Bujar Rauﬁ, Florije Ismaili, Jaumin Ajdari, Mexhid Ferati and Xhemal Zenuni

Faculty of Contemporary Sciences and Technologies, South East European University,

Ilindenska 335, Tetovo, Macedonia

Keywords:

Semantic Web, Linked Open Data, User Adaptive Software Systems, Adaptive Web-based Systems.

Abstract:

In recent years, a substantial shift from the web of documents to the paradigm of web of data is witnessed.

This is seen from the proliferation of massive semantic repositories such as Linked Open Data. Recommending

and adapting resources over these repositories however, still represents a research challenge in the sense of

relevant data aggregation, link maintenance, provenance and inference. In this paper, we introduce a model of

adaptation based on user activities for performing concept adaptation on large repositories built upon extended

generic ontology model for Adaptive Web-Based Systems. The architecture of the model is consisted of user

data extraction, user knowledgebase, ontology retrieval and application and a semantic reasoner. All these

modules are envisioned to perform speciﬁc tasks in order to deliver efﬁcient and relevant resources to the

users based on their browsing preferences. Speciﬁc challenges in relation to the proposed model are also

discussed.

1 INTRODUCTION

The expansion of the web in the recent years, espe-

cially with the proliferation of new paradigms such

as semantic web, linked semantic data and the so-

cial web phenomena has rendered it a place where

information is not simply posted, searched, browsed

and read. The web has been transformed into a place

where content and user experiences can be adapted in

various ways for a single purpose of action, which is

meeting user’s needs. This adaptation is being con-

ducted through Adaptive Web-Based Systems, which

tend to arrange its internal link structure, content, or

both, based on user browsing patterns.

User adaptive software systems in general and

adaptive web-based systems in particular have been

developed in two distinct pillars. One was particu-

larly focused on document space related to the area

of Personalized Information Retrieval(PIR) and the

other originating from the hypertext space in the

ﬁeld of Adaptive Hypermedia (AH) (Steichen, 2012).

PIR extends the classical view of one-size-ﬁts-all

paradigm by taking into account historical interac-

tions and ﬁnding most relevant documents for a sin-

gle query. On the other hand, AH approach tends to

ease this sort of search by providing the most relevant

browsing content and linking it with respect to a rich

representation of user characteristics. Such charac-

teristics could be preferences, history or prior knowl-

edge. While PIR is mostly query based, AH is mainly

browsing oriented.

An ongoing challenge is the semantic retrieval

and adaptation of resources over a huge repositories

such as Linked Open Data (LOD) as well as the fa-

cilitation of the collaborative knowledge construction

by assisting in discovering new things (Bizer, 2009).

Likewise, recommending and adapting resources over

a huge semantic repositories such as LOD (Linked

Data, 2013) still represents an open issue regard-

ing the relevant data aggregation, link maintenance,

provenance and inference. Rauﬁ et al. (2011) outlines

the importance of adoption of Semantic Web Tech-

nologies for content adaptation. The reason for such

adoption, which can be furthermore extended to big

resource repositories, can be summarized as follows:

1. Semantic web can be used to describe every doc-

ument or other resources in adaptive web-based

system’s repository (documents and any other

smaller units) according to a given vocabulary.

2. After their semantic description, these resources

can become machine processable and conceptu-

ally determinable.

3. These aforementioned resources are scaling op-

timally with no particular increase in processing

power. Such examples have been already tested

and are up and running in multiple Triple Stores

like: AllegroGraph, Stardog, OpenLink Virtuoso,

103

Rauﬁ B., Ismaili F., Ajdari J., Ferati M. and Zenuni X..

Semantic Resource Adaptation Based on Generic Ontology Models.

DOI: 10.5220/0005096001030108

In Proceedings of the 9th International Conference on Software Paradigm Trends (ICSOFT-PT-2014), pages 103-108

ISBN: 978-989-758-037-6

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

BigOWLIM, Garlik4store, Bigdata(R), YARS2,

Jena TDB, Jena SDB, Mulgara, RDF gateway or

Jena with PostgreSQL (Large Triple Store, 2014).

4. These semantically describable resources can be

further on queried through various endpoints and

presented to the user.

In this paper we introduce an architecture model for

semantic resource adaptation over large repositories

based on generic ontologies. The generic ontologies

in this model are considered those ontologies that can

be successfully retrieved and applied from the seman-

tic reasoner in order to perform the adaptation pro-

cess.

The rest of this paper is organized as follows: sec-

tion 2 introduces technology background related to

semantic web, section 3 describes some related work

introduced to the topic, section 4 elaborates the archi-

tecture of the proposed model for semantic resource

adaptation based on generic ontologies, it describes

its mechanisms through a simple use case and opens

up some interaction design challenges in relation to

resource adaptation and ﬁnally, section 5 concludes

this paper.

2 BACKGROUND

The main techniques used to represent the semantic

web approach in our proposed model is based on Re-

source Description Framework (RDF), RDF Schema

(RDFS) and Web Ontology Language (OWL) which

is portrayed as a standard in ontology modeling and

representation.

RDF is categorized as a general-purpose triple-

based language used for describing information on

the Web. RDF statements are deﬁned as subject-

predicate-object triples. From the perspective of

LOD, RDF offers ﬂexibility in the sense of publica-

tion, but creates a drawback in the point of view of

scalability (Fernandez, 2012). Flexibility of RDF is

seen through the aspect of resource exchange (many

proprietary formats such as RDF/XML (RDF/XML,

2004), Notation3 (N3) (Berners-Lee, 2011), Turtle

(Turtle, 2014) and RDF/JSON (RDF/JSON, 2013))

and resource publication (dereferenceable URIs, RDF

data dumps or SPARQL endpoints (SPARQL, 2011).

Drawback on the other hand is seen in the process of

resource consumption which is constrained by factors

such as serialization formats of RDF resources and

indexing of the same (Fernandez, 2012).

RDF Schema (RDFS) represents a semantic ex-

tension of RDF which provides mechanisms for de-

scribing groups of related resources (classes or prop-

erties) and the relationships between these resources.

RDFS allows extension of the classical triple based

RDF in the sense of well deﬁned classes and proper-

ties.

Web Ontology Language (OWL) is used to ren-

der the available resources to be machine processable,

where the meaning of each term and their relation-

ships are explicitly denoted. OWL has more extended

mechanisms for expressing meaning and semantics

than XML, RDF, and RDFS, thus OWL goes beyond

these languages in its ability to represent machine in-

terpretable content on the Web (Cristophides, 2008).

The important aspect related to ontologies is that

there are very solid reasoners or inference engines

available (HermiT, Fact++, Pellet, RacerPro etc.) that

can draw fair conclusions in huge semantic resources

at a reasonable time (Dentler, 2011). These vast re-

sources, however, pose a challenge in deciding the

most relevant resources for a given task. To address

this issue, we introduce a model of adaptation by

grasping user activities for performing concept adap-

tation on large repositories such as linked open data

based on extended generic ontology user model for

Adaptive Web-Based Systems initially introduced in

(Rauﬁ, 2009).

3 RELATED WORK

The undergoing research concerning adaptation of re-

sources in huge semantic repositories such as Linked

Open Data is focused around potential identiﬁcation

of complementary feasibilities between Personalized

Information Retrieval (PIR) and Adaptive Hyperme-

dia (AH) (Steichen, 2012). The abovementioned pos-

sible complementariness gives various ways for hy-

brid approaches which allow techniques in relation to

retrieval process such as: query adaptation, adaptive

retrieval and adaptive composition and presentation.

In query adaptation techniques, the focus is given

mainly on the process of query disambiguation, per-

sonalization and result diversiﬁcation. All this in re-

lation to medium and small closed corpus semantic

repositories and no user model is available. Adapta-

tion technique is focused around semantic reasoning,

query expansion or substitution. An example of this

group is seen in (Chirita, 2007) where query adapta-

tion is performed through the process of query expan-

sion with new terms depending on the user experience

(novice or experienced visitors) or query ambiguity.

Concerning adaptive retrieval techniques, the cen-

ter of attraction is given mainly to content-concept

mapping around small open/closed corpus domain,

user model represents an overlay knowledge model

based on preferences and adaption algorithms are

ICSOFT-PT2014-9thInternationalConferenceonSoftwareParadigmTrends

104

mainly rule based. One promising semantic adapta-

tion techniques lies in the ﬁeld of Open Hyperme-

dia Systems (Wiil, 2006). Even though, OHSs do

not provide direct adaptation, they can be combined

with such systems to provide various personalization

strategies. One approach of such model is elaborated

in (Fulantelli, 2000).

In adaptive resource composition and presenta-

tion, a more distinctive grouping between informa-

tion retrieval and adaptive hypermedia techniques

is evident. This distinction is seen from the per-

spective of adaptive behavior, user involvement and

modeling, adaptation algorithms and their scalabil-

ity. In this sense, techniques such as: faceted rank-

ing search, adaptive navigation, user intent elicitation,

rule based and semantic reasoning for adaptive com-

position from one side and adaptive summarization,

link annotation, cluster selection, preference summa-

rization for adaptive presentation from the other are

available (Steichen, 2012).

It is worth mentioning that partial solutions of hy-

brid systems that tend to fuse the approaches of PIR

and AH are available. One such solution is given in

(Shekhar, 2013) where the application is consisted of

three modules: web extraction module which contains

the user browsing history, the semantic knowledge

base module which contains the ontology in accor-

dance with the user’s browsing history and ﬁnally the

reasoner module which parses through the ontology

and generates a result based on a search query. An-

other methodology is seen in (Bielikova, 2006) and

(Bielikova, 2008) which tends to model the content

of adaptive web application based on ontology. The

ontology is modeled from a domain model and a do-

main dependent part of a user model. While the for-

mer work focuses more on content modeling using

an ontology, the latter tends to design the ontological

representation of a model which is suitable for further

reuse.

Main drawback of the above elaborated ap-

proaches lies mainly in their closed corpus nature,

considering that, to the best of our knowledge, they

have not been designed and tested on large reposito-

ries such as the Web of Data. In the section that fol-

lows, we elaborate a model designed speciﬁcally to

address the complexity and vastness of large seman-

tic repositories with a particular emphasis on Linked

Open Data.

4 PROPOSED ARCHITECTURE

In order to address the issues risen above, a hybrid

architecture that focuses on the issues of both per-

sonalized information retrieval in semantic level and

resource adaptation is proposed. The architecture is

envisioned of four main modules with speciﬁc tasks.

The semantic resource adaptation architecture is illus-

trated in Figure 1 and is consisted of the following

elements:

User

Browser

User Data Extraction

User Knowledgebase

Ontology Retrieval

and application

Semantic Reasoner

The Semantic Repository

Figure 1: The architecture of a semantic resource adapta-

tion.

User data extraction module will extract the user

browsing activities and insert them to the user knowl-

edge base (ontology) as object instances or data prop-

erties. User browsingactivities are kept in various for-

mats (logs, database tables, etc) on the browser side

and requires additional intermediary tasks for such

extractions and data mappings.

User knowledge base represent an extended ver-

sion of the ontology initially introduced in (Rauﬁ,

2009) that depicts user activities on the semantic

level. These activities usually comprise user sessions,

visited content and generated user views as part of the

adaptation. Browsing activities from the data extrac-

tion phase given earlier, are also stored on the ontol-

ogy in the form of object instances or data properties.

Ontology retrieval and application is a phase

of ontology aggregation and fetching from various

repositories based on user browsing behavior. For in-

stance, say a user is interested in a particular topic,

and in order to adapt to that concrete user activity,

a speciﬁc ontology from publicly available reposito-

ries (Cupboard, Knoodl, Schemapedia, SchemaWeb,

TONES, etc.) should be acquirable.

Finally, the reasoner module makes it possible the

connection between the previous modules to work to-

gether in the background. The reasoner draws con-

clusions on instances and data properties based on the

SemanticResourceAdaptationBasedonGenericOntologyModels

105

classes, class hierarchies and various restriction prop-

erties deﬁned explicitly in the ontologies. The im-

portant part of the architecture lies in the user model

deﬁned as the user knowledge base and the ontology

retrieval as depicted in Figure 1 above.

The steps involved in the creation of the ontology

comprises of the following aspects:

• Deﬁnition of concept classes and hierarchies

• Deﬁnition of concept properties

• Deﬁnition of concept relationships and

• Deﬁnition of concept instances and data proper-

ties

Main concept classes in the user knowledge base on-

tology are as follows: User, Session, Preferences,

UserView, Content, Page and AtomicUnit. Each of

these classes are hierarchically deﬁned in a parent-

child relationship. To summarize shortly, a user can

have a particular session, preference and a view which

in turn is consisted of a resource content such as pages

or more ﬁne grained elements we coined as atomic

units (Rauﬁ, 2009). All of the above entities are re-

lated at semantic levels through concept relationships

such as: hasSession, hasPreference, hasUserView etc.

The complete ontology and their relationships is

depicted as in Figure 2.

User

Preferences

Content

Page

Atomic Unit

Link

Position

Type

UserViews Session

hasPreference

hasSession

hasUserView

hasContent

hasPage

hasAtomic

belongsTo

hasType

hasPosition

isLink

Figure 2: The User Ontology.

During ontology retrieval and application phase,

the preferences class from user ontology given above

serves as an input to ontology retrieval. After the

ontology is retrieved, a proper class hierarchy, do-

main attributes as well as datatype identiﬁcation is

performed through ontology parsing. Finally, the

retrieved elements are returned to Content class in

user knowledge base which will serve as inputs for

UserView generation.

The detailed view of ontology retrieval, parsing

and application is illustrated in Figure 3. The dashed

lines illustrate the process of resource return to Con-

tent class of user knowledge base ontology.

Preferences

Content

...

User Ontology

Rerteiving and Parsing

Ontology Retrieval

and Application

Reasoner

Inference

Figure 3: Detailed view of ontology retrieval and applica-

tion.

4.1 Semantic Resource Adaptation: A

User Scenario

John is visiting Berlin for the ﬁrst time. He is using

his mobile device to explore interesting sites around

the city. A friend has recommended him to visit Re-

ichstag’s building, which he types in his semantically

enabled mobile web browser. The browser shows im-

portant details about Reichstag’s building along with

a link ’Directions’, which shows John how to get to

the building (the browser does this based on John’s

smartphone geolocation). John decides to click on the

link and follow the directions. Leveraging geoname

(GeoNames, 2014) information, the browser suggests

John to also visit Jakob Kaizer Haus (the browser now

suggests other important sites following John’s trail

to get to Reichstag’s building). John initially did not

know anything about this building, but since it is on

his way, he is glad to explore it.

The typical course of events for this particular user

scenario seen from systems perspective involves the

following steps:

Step 1: User visits a particular semantically capa-

ble web page.

Step 2: User wants to know more for a particular

resource.

Step 3: The application checks for other relevant

resources based on retrieved ontologies and de-

rived inference.

ICSOFT-PT2014-9thInternationalConferenceonSoftwareParadigmTrends

106

Step 4: These resources, in the form of high-

lighted content or other visually aided methods is

presented to the user.

Alternate courses of events might also involve:

Step 5: User browsing patterns are highly id-

iosyncratic so that a proper preference and conse-

quently, a relevant resource cannot be identiﬁed.

Step 6: In case of step 5, no adaptive resource is

delivered.

The above elaborated use case also implies several

user interaction challenges that also need to be ad-

dressed. These challenges can be viewed from the

perspective of resource entities, user focus preserva-

tion and user friendliness.

4.2 User Interaction Challenges

The key beneﬁt of large semantic repositories such

as linked open data seen from user’s perspective men-

tioned above,is the arrangement of integrated data ac-

cess from heterogeneous data sources which initially

might not be accessible to users (Bizer, 2009). In

this direction the following several challenges could

be identiﬁed:

Entity-centered User Interface: In a typical web

browsing scenario users navigate from page to page,

which represents the page-centered navigation. This

type of navigation, however, is not suitable for LOD

as it does not leverages the full potential of these large

semantic repositories. To offer a heightened user ex-

perience a new paradigm of entity-centered naviga-

tion should be adopted. Such, navigation will allow

a ﬁne-grained interaction between entities that will

generate a more speciﬁc entity to entity access com-

pared to page to page. In order to offer the best user

experience, the entity-centered navigation should al-

low bidirectional browsing between underlying enti-

ties. For instance, referring to the above mentioned

scenario the user moves from Raichstag subject entity

to Jakob Kaizer Haus object entity connected through

a particular predicate like e.g. onTheWaySites. De-

spite that such scenario should be unambiguous from

the machine processing point of view, it must be nat-

ural and cognitively plausible from user experience

perspective.

Maintaining User’s Focus: When navigating

large semantic repositories, users could lose their fo-

cus from the amount of options offered. Maintain-

ing user’s focus based on the selected preferences is

a challenging task. For instance, while John, on his

way to Reichstag’s building he should be offered with

other relevant sites on his way to destination.

User Friendly Information Representation:

The LOD functionson the principle of displaying data

organized in triplets: object-predicate-subject. Typi-

cally, such information is shown as it is without guid-

ing the user through the process of information dis-

covery. While such navigation is effective, it under-

mines the serendipity of information, which results

in low user experience. For instance, when John

searches for Reichstag he is offered with directions

link, which depicts how the interface should antici-

pate user’s possible future interests.

Some aspects of the issues presented are partially

addressed in (Catasta, 2009) through Sigma search

engine. However the approach gives an insight of the

limits in capabilities and functionalities of the system

when huge repositories are the case.

5 CONCLUSIONS AND FUTURE

WORK

In this paper, an architecture of semantic resource

adaptation on massive repositories is introduced. The

system tends to adapt its content based on user pref-

erences, retrieved external ontologies and reasoning

mechanism. The process of adaptation is done en-

tirely through reasoners and reasoning techniques

taken from semantic web technologies. The fetched

resources are returned to user ontology as content and

proper user views.

The future work would involve:

• Development on speciﬁc parts of the architecture

modules and implementing the system as a whole.

Some part of the architecture elaborated above are

implemented such as the user knowledgebase as

given in Figure. 2 and some initial results regard-

ing reasoning on that knowledgebase.

• Exploring various ways of reasoning mechanisms

and reasoners against our architecture. Compar-

ison analysis between reasoners such as Pellet,

Fact++ and Hermit on one hand and SWRL rule

language on the other hand would be interesting

from the sense of retrieved resources relevant to

the user.

• Evaluation of the aforementioned proposed archi-

tecture against these large linked repositories from

the perspective of its efﬁciency, resource recall

and precision. One particular interesting evalu-

ation would be from the perspective of scalabil-

ity having in mind that we are dealing with huge

amounts of data.

• Investigating best interaction modality to provide

the retrieved resources to users. This will involve

SemanticResourceAdaptationBasedonGenericOntologyModels

107

experimenting with various adaptive interfaces to

offer users rich user experience.

If these future task are to be addressed properly, we

expect that the proposed architecture will substan-

tially facilitate the access to linked open data as well

as realize the full potential of the web not only as doc-

ument but also as a data information space.

REFERENCES

Bizer, C. Heath, T., Berners-Lee, T. (2009). Linked data-the

story so far. In International Journal on Semantic Web

and Information Systems (IJSWIS). 1-22. 5/3

Berners-Lee, T., Conolly. D. (2011). Notation3 (N3): A

readable RDF syntax. W3C Recommendation [On-

line]. http://www.w3.org/TeamSubmission/n3/ Ac-

cessed: April, 2014.

Bielikova, M., Moravcik, M. (2008). Modeling the

Reusable Content of Adaptive Web-Based Applica-

tions Using an Ontology. In Advances in Seman-

tic Media Adaptation and Personalization Studies in

Computational Intelligence. Vol. 93. pp. 307-327

Bielikova, M., Moravcik, M., (2006). Modeling the Content

of Adaptive Web-Based System Using an Ontology.

In First International Workshop on Semantic Media

Adaptation and Personalization, SMAP ’06. vol., no.,

pp.115-120

Catasta, M., Cyganiak, R., Tummarello, G. (2009). Towards

ECSSE: live Web of Data search and integration. In

Proceedings of the Semantic Search 2009 Workshop

at WWW2009.

Chirita, P., Firan, C. S., and Nejdl, W. (2007). Personalized

query expansion for the web. In Proceedings of the

30th Annual International ACM SIGIR Conference on

Research and Development in Information Retrieval.

SIGIR ’07, pp. 7-14

Cristophides, V., Collard, M., Gutierez, C. (2008). Seman-

tic Web, Ontologies and Databases. In VLDB Work-

shop, SWDB-ODBIS. Lecture Notes In Computer Sci-

ence (LNCS). 5002

Davis, I., Steiner, Th., Le Hors, A.J. (2013) RDF 1.1 JSON

Alternate Serialization (RDF/JSON). W3C Recom-

mendation. 2013

Dentler, C., Cornet, R., Teije, T.A, Keizer, N. (2011). Com-

parison of Reasoners for large Ontologies in the OWL

2 EL Proﬁle. In Journal of Semantic Web. (1) 1-5.

IOS Press.

Fernandez, H. D. (2012). Binary RDF for scalable publish-

ing, exchanging and consumption in the web of data.

In Proceedings of the 21st international conference

companion on World Wide Web (WWW12). 133-138.

Fulantelli, G., Rizzo, R., Arrigo, M., Corrao, R. (2000).

An Adaptive Open Hypermedia System on the Web.

In Proceedings of the International Conference on

Adaptive Hypermedia and Adaptive Web-Based Sys-

tems (AH ’00). Brusilovsky, P., Stock, O. and Strap-

parava. C. (Eds.). Springer-Verlag, London, UK, UK,

305-310.

GeoNames geographical database. (2014). The

GeoNames geographical database. [Online]

http://www.geonames.org/ Accessed May, 2014

Rauﬁ, B., Luma, A., Zenuni, Xh., Ismaili, F. (2011). Se-

mantic Adaptation Approach for Adaptive Web-Based

Systems. In Proceedings of the 2nd International Sci-

ence and Technology Conference (ISTEC’2011). 705-

708. Istanbul, Turkey

Rauﬁ, B. Ismaili, F. Zenuni, X. (2009). Modelling a com-

plete ontology for adaptive web based systems using a

top-down ﬁve layer framework. In Proceedings of the

31st International Conference on Information Tech-

nology Interfaces (ITI 2009). pp. 511,518, 22-25 June

2009. Dubrovnik/Croatia

Shekhar, M., Saravanaguru RA. K. (2013). Semantic Web

Search based on Ontology Modeling using Protege

Reasoner. arXiv preprint arXiv:1305.5827.

Steichen, B. Ashman, H. Wade, V. (2012). A compara-

tive survey of Personalised Information Retrieval and

Adaptive Hypermedia techniques. In Journal of Infor-

mation Processing and Management. 48, 4, pp. 698-

724.

SPARQL 1.1. Query Language. (2011). W3C Recommen-

dation. 2011

RDF/XML Syntax Speciﬁcation (Revised). (2004). W3C

Recommendation. 2004

Turtle - Terse RDF Triple Language. (2014). W3C

Recommendation. [Online] http://www.w3.org/

TR/2014/REC-turtle-20140225/ Accessed April,

2014

Linked Data. (2013). [Online] http://www.w3.org/ stan-

dards/semanticweb/data. Accessed April, 2014

Large Triple Store. (2014). [Online] http://www.w3.org/

wiki/LargeTripleStores Accessed April, 2014

Wiil, U. (2006). Open Hypermedia: Systems, Interoper-

ability and Standards. In Journal of Digital Informa-

tion. 1(2). Retrieved from http://journals.tdl.org/jodi/

index.php/jodi/article/view/9/9

ICSOFT-PT2014-9thInternationalConferenceonSoftwareParadigmTrends

108