Content uploaded by Janneth Chicaiza
Author content
All content in this area was uploaded by Janneth Chicaiza on May 05, 2020
Content may be subject to copyright.
Recommendation of Open Educational Resources.
An approach based on Linked Open Data
Janneth Chicaiza, Nelson Piedra and Jorge Lopez-Vargas
Departamento de Ciencias de la Computaci´
on
Universidad T´
ecnica Particular de Loja
Loja 1101608, Ecuador
Email: [jachicaiza, nopiedra, jalopez2]@utpl.edu.ec
Edmundo Tovar-Caro
Departamento de Lenguajes y Sistemas Inform´
aticos
Universidad Polit´
ecnica de Madrid
Madrid, Spain
Email: etovar@fi.upm.es
Abstract—In an open and distributed platform as the Web
there are problems associated with the heterogeneity and in-
formation overload. In this context, teaching community and
learners may need some support to discover the Open Educational
Resources best suited for their learning processes. In this paper,
the authors propose the high-level design of a framework for the
recommendation of learning content in a flexible way. In order
to provide a personalized set of resources, an adaptive approach
of filtering is used according to the data available of each user.
To achieve this goal, the framework has been designed taking
into account the best features provided by technologies of the
Semantic Web in order to find online material.
I. INTRODUCTION
Open Educational Resources (OERs) have the potential to
encourage self-learning and life-long learning. However, users
can experience barriers to find the right material.
Due the large amount of available open content on the Web,
teaching community and learners may need some support to
discover the OERs which can be adapted to their own learning
environment.
Google is the most popular tool to find material in the
Web. However, this kind of systems - known as general search
engines - don’t leverage the user’s knowledge, so thousands of
results would be returned to people. A recommender system
unlike the search engine takes into account the users interests,
moreover it is designed to give a better support to user during
a session of searching.
An effective recommender system requires big amounts
of data [1]. In closed corpus or institutional virtual learning
environments, this requirement could be achieved because data
could be obtained from their repositories. However, in the Web
- environment underlying to OERs - [2] this is a big issue
to address, since in open repositories and services it is not
possible to control the quantity and the quality of the resources’
and users’ metadata.
In last years, information retrieval and recommender sys-
tems are including semantic capabilities in order to reduce
problems associated to information overload and increase their
performance, i.e. technologies of the Semantic Web enable
designing a new generation of tools for discovering resources
on the Web.
In the following section, the challenges related to OER rec-
ommend process are discussed in the context of open learning.
Based on the limitations related to the current proposals, in this
research, the main problem is presented.
In section III, in order to improve the OER localization, a
recommendation framework of OERs is presented. It leverages
structured and organized data available in open knowledge
systems, these data are used to describe OERs and improve
their discoverability by machines. Another feature of the pro-
posal is the use of open vocabularies and formal languages to
improve the reuse and interoperability of data between several
applications. The process of data filtering is implemented
by a query engine, which selects the material that might be
interesting to an user.
Finally, in the section IV, the conclusions and future work
are presented.
II. THEORETICAL BACK GROU ND A ND REL ATED WO RK
In environments with information overload, two approaches
are used to find content: search engines and recommender
systems. Although they can be combined to work together,
they often work independently since each one can be more
effective in certain settings.
A search system is generally used when user knows what
he wants. In contrast, a recommender system is more efficient
when user can’t accurately describe his information needs, i.e.
it is oriented toward situations that are less known by the user
[1].
In order to help the user to find the most relevant OERs
according to his preferences, this paper present an approach
based-on knowledge filtering which is flexible to support the
needs of heterogeneous or anonymous users.
A. Introduction to filtering information systems
Information filtering systems have become popular on the
Web, especially in commercial field. One of the most popular
services is offered by Amazon. User of the company’s website
receives similar products to the item that he is watching [1]. If
the user decides to select one of the suggested items, he would
receive more information about it or he can interact with them
in any way [3].
Recommender systems are based-on information filtering
methods to predict the elements that will be recommended
to a user. The concept of ”filtering” determines the selective
nature of a recommender system. According to [4], using right
techniques and enough information, these kind of systems
make the decision whether a document may be relevant or
useful to a particular user.
The filtering is performed based-on preferences of individ-
uals or groups of users [4]. The system identifies preferences
taking advantage of traces that the user leaves when interacts
with it [1]. The information about the visited pages is usually
leveraged for calculating the items’ rating seen by a user.
To generate recommendations there are three generic ap-
proaches: based on content, collaborative and based on knowl-
edge. Each approach applies a specific process to produce
results[5].
Filtering based on content requires processing the items’
content or their metadata. It is based on the fact that items
with similar characteristics will be rated similarly by a group
of users.
On the other hand, collaborative filtering uses the notion
of community of users. It is based on the fact that people with
similar tastes or preferences will rating the same elements on
similar way [6]. In this case, it is not necessary to access the
items’ content. Rather, a corpus of material rating according
to user preferences is required.
Finally, the third method of filtering leverages knowledge
of users and items in order to produce recommendations
according to the available data [7].
Unlike systems based on content and collaborative systems,
knowledge based ones are prepared to take advantage of
heterogeneous repositories of open data freely available on
the Web [7]. These additional datasources are used to enrich
actual descriptions of resources and users, thus dealing with
the scarcity of rating data.
B. Recommender systems in OER context
In business, recommender systems has become popular
because of users interact by rating the products seen or pur-
chased. The amount of reviews and ratings is enough in order
to guarantee the quality of recommendations, in particular, a
pure collaborative system requires significant amounts of user’s
ratings [1].
In addition, classic recommendation approaches, based on
content and collaborative systems, have the issue known as
cold start problem which occurs when new users or new
elements are added to the system. New users can’t receive
recommendations if they don’t rate any material. On the Web,
as a platform for open learning, there are users with anony-
mous profile, who will be exempt of getting recommendations
if a classical approach is adopted [7].
In closed or specific e-learning environments, recommen-
dation can be performed from a classic approach of filtering.
Data necessary for this task are collected from the reposito-
ries of institutional learning systems. The recommendation of
learning objects (LOs) is performed by comparing the student’s
profile and the LO’s metadata, which are generally described
by standards such as LOM or Dublin Core. Due to the open
nature of the OER movement, in the Web, it is not feasible to
use a single metadata standard. In addition, in OER context,
data about resources could be missing. The lack or shortage of
data can affect the performance of recommender systems [7].
To obtain user’s data is also a complicated task, in an
open learning environment, because of learners don’t enroll
to a specific institution and the underlying learning platform
so there is little knowledge of the user or it is distributed
in different web services. Therefore, recommender systems
must incorporate new capabilities to build the student profile
dynamically by enriched it with external data and by offering
the support for the user to support this process on a natural
way.
Another difference between recommendation focused on
formal learning vs. learning based-on OERs is that in the
first case there are user’s profiles well defined. Conversely,
in open educational systems or in informal learning there are
broad spectrum of users with different needs of information
and learning. Therefore, the discovery of educational materials
should go beyond finding a few specific documents [8], rather
the user could require assistance to find relevant material.
To achieve this aim, the system should include capabilities
of dialogue and assistance to user to explore resources [1].
The conversational ability of a recommender system is key in
open online learning environments, especially when there are
anonymous users.
Finally, the OER location is a difficult task due to there
isn’t a common knowledge system to organize them. Currently,
many OERs are not classified or are based on different thesauri
where equivalent topics are in different languages or they
are written according to different vocabularies and levels of
granularity. In a report published in 2013, users replied that
one of the key aspects is to improve the organization of courses
according to disciplines of knowledge [9].
For all the above, the recommendation focused on OERs
should handle issues from the lack, sparcity, distribution and
heterogeneity in data of users and resources. Therefore, new
approaches must be designed to identify learning resources
useful for all types of users in open environments.
C. Related Work
Specific proposals focused on the recommendation of OER
have not been identified. Rather, some initiatives designed
and tested in closed corpus of material or institutional virtual
learning environments (VLE) have been found in [10], [11],
[12].
The common behavior of the content-based filtering pro-
posals consists in creating explicitly some metadata and rela-
tionships between learning objects. In some cases each object
of the collection is cataloged manually according to a particular
scheme of classification of knowledge. On the Web, as an open
learning platform, although agreements could be achieved, it
is not coherent to control the use of certain metadata standards
or knowledge schemas to describe materials.
Another common feature that has been detected in the
reviewed studies is that their evaluations were carried out on
closed platforms and focused on the formal learning scope.
Therefore, it is not assured the scalability and flexibility in
less structured and unconventional environments as the OERs
[13].
As a result of this analysis is possible to claim that most
efforts of filtering information have mainly focused on: i) how
to obtain and model the user’s preferences, and ii) identifying
algorithms to predict items that could be positively rated by a
user. However, as stated in [14] less attention has been paid
to the decision-making processes of whom use these systems.
Proposals such as [12], [11] have endeavored by au-
tomating the recommendation process, when a main feature
of the system should be to help people so they can make
good decisions. To meet this objective, the system must be
conversational and allow the participation of users when they
needed.
Recommender systems based on knowledge are ideal to
support decision-making [14]. Knowledge related to decision
making becomes a success factor in the implementation of
recommendation systems. It’s important find the right balance
so that the orientation is not too detailed or too intensive
especially for those who require greater autonomy during the
localization process of educational material [15].
In this research, a filtering approach based-on knowledge to
recommend OERs has been adopted, specifically, a framework
based on Linked Data and Semantic Web technologies is
designed. From the technological point of view, a knowledge-
based filtering mechanism is more suitable for OER context
because: (i) reduced vulnerability to missing data; (ii) ability
to leverage open and heterogeneous knowledge sources; and
(iii) greater flexibility to include user’s support and feedback
features.
Recommender systems based on knowledge are oriented of
two lines of work: i) use of ontologies and RDF data available
on open sources in order to leverage semantic content, and ii
) definition of recommendation algorithms suitable for linked
data.
Regarding to the first group of works, an ontology can
be used to measure correlations between resources and to
infer semantic characteristics from ontologies. Once users
and/or resources to recommend have been described by an
ontology, the underlying semantics is used to find better results
[16]. The inference of semantic features is made possible by
applying ontological reasoning. Through domain ontologies,
new information about resources, users and even on the context
is discovered.
Other job in above line is presented in [17], that uses
domain inferences to fill user’s models with additional assump-
tions about the user’s interests by making inferences based on
the hierarchical structure of interests.
Regarding to the second group of approaches, new or
improved algorithms of recommendation have been proposed.
In [18] presents RecSPARQL which adds to SPARQL some
features of recommendation based on similarity between users
and resources independently of domain of interest.
In [19] a recommendation system for the musical field is
proposed. DBpedia is used to calculate the semantic distance
between bands and musical artists. The system called, dbrec,
was one of the first systems using open data sources such as
DBpedia. This paper analyzes incoming and outgoing links of
each artist, however, it does not exploit the existing semantic
relationships between musical artists described in the DBPedia
[20].
Unlike [19] in [21], the properties from DBpedia and
Linked-MDB1are extracted to make it possible the semantic
expansion of musical items descriptions, in this way is possible
to add a learning ability of the users profiles [20].
Another work using DPBedia is presented in [20]. In this
case, the data model is used to capture the complexity relation-
ships between users, items and other entities. Traversing paths
between users and items, recommendations are calculated
through a learning algorithm called SPRank classification.
The contributions made by the authors of [19], [21], [20]
are not precisely in education field, but they provide a basis for
asserting that a filtering mechanism based on linked data and
based on open knowledge schemes can be a good combination
to carry out a proposal in the field of open and informal
learning and specifically in the context of OERs.
III. DES IG N AND PR E-VAL IDATIO N OF TH E
RECOMMENDATION FRA MEW OR K
A. General Approach
In order to achieve that OER recommendation has a signifi-
cant impact on a user, a subset of most relevant resources must
be selected according to user’s interests expressed as subjects
or pieces of knowledge.
Concretely, the design of a framework for the personalized
location of OERs is proposed. The design consists of a
recommendation cycle based-on Linked Data and hierarchical
structures of knowledge. Figure 1 shows the flow of data that
is generated by the process of recommendation and its relation
to data models managed by the framework.
The recommendation cycle manages the data that charac-
terize learning resources and Web users. To make possible
the recommendation, the data must be encoded in machine-
processable formats. Currently, in the Web can be found large
amounts of data and semantic descriptions.
Data models are system components that define the struc-
ture to describe resources and users, and provide the means to
connect both representations and support different mechanisms
of information filtering.
From a technological point of view, the framework has
been designed under four principles:
1) Flexible. The filtering strategy must be applied ac-
cording to the available information.
2) User centered. The knowledge about the user must
lead the discovery of open educational resources. In
addition, user’s participation during formulation of
his queries should be listened by system so improving
its awareness on which mechanisms of support or
assistance provides according to level of knowledge
of an user about a topic.
3) Interoperable. Partial results during each stage of
recommendation must be able to be combined with
other strategies of recommendation and should be
able to be integrated by different applications.
1http://www.linkedmdb.org/
Fig. 1. Top-level view of the Recommendation Framework
4) Able to learn: The system must improve its perfor-
mance from datasets that people share and organize
in the Web. Concretely, these data should be used
with the following purposes: i) improve knowledge
about the users domain of interest; ii) enriching the
inputs and outputs of the system; and iii) determine
the best context of search.
To ensure that the solution is consistent with each of the
mentioned guidelines, four technology strategies have been
selected in its design.
•Filtering based on linked data. An open multidi-
mensional filtering mechanism allowing to support
heterogeneous scenarios.
•Enriching data from open sources structured. The
underlying knowledge to an system can be exploited
for different purposes, such as guiding the search by
facilitating the exploration between domain topics or
concepts, by improving results when user search or
filter the results according to a context.
•Semantic representation. To describe educational re-
sources and users, representations based on open vo-
cabularies have been selected. Content described by
semantic technologies facilitates its discovery [22].
•Scalable and reusable architecture in an academic
setting. A service-oriented infrastructure provides easy
and flexibility of integration with other systems and
maximizes the reuse of the individual components of
the framework in other solutions.
•Semi-automatic process according to a conversational
behavior. The user who receives the recommendations
can participate in each of the stages of recommenda-
tion process.
Then, more details of the processing cycle and the three
data models of the framework are provided.
1) Data Models: Three data models are needed to support
recommendations: i) to describe the users profile, ii) to rep-
resent OER, and iii) an intermediate scheme to connect both
representations.
In order to ensure the successful of a recommendation
model, a suitable user’s data model is required according to
the challenges related to OER context. In a previous work, an
open linked vocabulary is proposed to describe user’s profiles
of the open educational resources [23] .
Regarding of the OERs representation, they must be de-
scribed, classified and characterized to facilitate their discov-
ery by software agents. The characteristics of an OER are
grouped in a profile that complements the vocabulary proposals
LOCWD defined in [24].
Finally, as an intermediate scheme to connect the represen-
tations of users and learning resources it has been designed a
model based on knowledge organization systems (KOSs). The
intermediate layer of data is constructed through semantic en-
richment of a taxonomy or hierarchical structure of knowledge
from an open source of knowledge. The creation process of
this repository is detailed in [25]. The semantic enrichment
can leverage the dynamic and the flexibility representation of
knowledge in an open system, to feed and enrich the define
structure by a rigid scheme as a controlled thesaurus.
2) Data Processing Cycle: The data processing cycle con-
sists of four general phases:
1) A. Data Management: includes capturing metadata
from different repositories and Web sources and the
respective data generation depending on open vocab-
ularies. This stage ends with the storage of these data
in a local repository.
2) B. Enrichment: organizes and complements OER data
profiles as well as the users based on data published
in open structured or social sources.
3) C. Exploitation: this generic phase of implementation
of knowledge can be made with different purposes.
In this case, it focuses on the filtering information.
The implementation of the filtering mechanism can be
accomplished through various methods and technolo-
gies. Although given the ease of access to data from
open knowledge sources through query languages
such as SPARQL, the proposed filtering is based on
a query engine which selects resources according to
a set of criteria of user’s interest.
4) D. Feedback: incorporates different inference strate-
gies and semantic derivation on interaction data gen-
erated by users when interacting with the system. As
shown in Figure 1 this phase is cross since it is
intended for user intervention at any stage.
B. Application scenarios
The framework has been designed based on a processing
cycle composed of different phases and well-defined functions,
this feature determines the open nature of the approach.
Therefore, different application methods can be implemented.
The development of a application scenario may require
a different and partial implementation strategy of the sys-
tem. That is, depending on the purpose and context of the
recommendation is necessary to perform certain activities of
the cycle. In addition, the implementation of search tasks
and recommendation must be made based on the available
information, as well as users and OERs.
One option is to build all system, as a result, different users
could access to a OERs recommendation platform which could
be customized according to the interests, knowledge level and
learning objectives.
Another way is to choose certain functions of the process-
ing cycle and implement them as independent Web services.
Thus each function could be integrated in hybrid recovery or
filtering information solutions.
According to this second approach, in order to improve the
discovery of the published materials, an automatic classifier
has been implemented. It organizes OERs according to the
defined disciplines by a thesaurus [25].
Other possible application of this proposal is to support
the OERs recommendation according to the metadata of the
materials and to the available users profile.
In order to explain the potential of the proposal to support
different location scenarios of material, a implementation strat-
egy of the framework is detailed. It was designed to find OERs
according to the concepts of interest of anonymous users.
To support this method, two data sources are required: i)
one RDF dataset of concepts or entities associated with each
learning resource, and ii) the user profile defined in function
on certain attributes such as knowledge areas, topics or specific
entities of their domain of interest.
Regarding to the processing tasks, three phases of data cy-
cles are essential: the acquisition of data from selected sources,
enrichment of OERs data through thematic classification and
exploitation of data through a filtering mechanism based in
queries.
To illustrate the recommendation mechanism is considered
as input sources: i) a profile of prefabricated users, ii) the RDF
dataset about OER data, LOCWD2generated in a previous
work [24], and iii) the knowledge base created from semantic
annotations recognized in the content of OERs [25].
As for the user’s profile, as a illustrative case it has been
defined a model based on the preference of two semantic
concepts: Java (programming language) and the knowledge
field from UNESCO, Programming Language. All OERs topic
having the same resources as defined in the user profile will
be recommended to the user.
The query to identify OERs that fits the users predefined
interests is presented in the following script. The results are
sorted according to weight or strength of the relationship
between OER and every field of knowledge that is part of
the user’s preferences.
2http://serendipity.utpl.edu.ec/sparql
PREFIX dc t : <h t t p : / / pu r l . o r g / dc / t e r m s />
PREFIX sch ema : <h t t p : // sc he m a . o r g/>
PREFIX rs d : <h t t p : / / r e c s y s . u tp l . ed u . ec :8 0 8 0 / o e r / r e s o u r c e />
PREFIX rs o : <h t t p : / / r e c s y s . ut p l . e d u . ec : 8 0 8 0 / oe r / sc he m a/>
SELECT DISTINCT ? wei g h t ? o e r
WHERE
{
VALUES ?interest {f oa f : i n t e r e s t }
rs d : u s e r0 2 ? in t er e st ? us e rI n t er e s t ; r s o : m a i n I n t e r e s t ? us e r M a i n I n t e r e s t .
? o e r r d f : t y p e <h t t p : // p u r l . o rg / l oc wd / sc he ma #OER>.
? oe r d c t : su b j e c t ? u s e r I n t e r e s t ; fo a f : t o p i c ?o e rT o p ic .
? o er T o p ic r s o : s u b j e c t ? us e r M a i n I n t e r e s t ; rs o : w e ig h t ? we i g ht .
}
GROUP BY ? o e r
ORDER BY de s c (? wei g h t )
With this kind of queries, the semantic representation of the
user is compared with the set of annotations that each OER
has in the knowledge base.
In order to check the system’s ability to provide customized
sets of OERs according to the user profile, the characteristics
of the provided results are compared by the proposal system
against the results returned by traditional metric system used in
filtering systems based on content. Concretely, we used the LSI
algorithm (Latent Semantic Indexing), which is implemented
by Gensim3, library of the Python.
Table I shows the Top-ten results obtained by the proposed
model based on linked data (LD) and the respective positions
that every OER recommended receives according to the LSI
model.
TABLE I. TOP -TE N LI ST O F TH E OERS
Id OER 756 3362 2615 3365 1607 3384 513 2617 576 1702
LD 1 2 3 4 5 6 7 8 9 10
LSI 11 14 17 33 2 25 26 27 10 51
Analyzing the ranking of the top-ten OERs recommended
by the LSI model, approximately the 50% of the recom-
mendations do not appear among the results generated by
LD. The omitted OERs correspond to resources that talk
about programming and programming languages but they not
necessary talk about JAVA as programming language. The
results obtained in this case confirm that the system based in
queries is effective to exclude results that do not fit the user’s
profile.
IV. CONCLUSION
In open learning systems, the spectrum of users accessing
the Web to finding the right material to support their activities
is significative. Anonymous users are one of the most repre-
sentatives. In OER context, there are different motivations for
which users search this type of material. Depending on the
user’s profile -teacher, student or selflearner-, the information
needs may change, therefore, motivations and search intention,
in each case, may be different. This feature makes the task of
recommendation a challenge.
In addition to the large amount of resources and data on
the Web, the heterogeneity of content and the almost non-
existent management of semantics of traditional systems make
more difficult the task of find open learning resources. In
this context, conventional techniques of recovery and Web
information filtering are limited.
In this paper, authors have proposed a framework oriented
to the OER location according to preferences of user’s profile.
3https://radimrehurek.com/gensim/
The underlying knowledge to a domain created and organized
to people has been used to guide the recovery of material in
different contexts.
From the technological point of view, the filtering method
is driven by a knowledge-based approach and specifically is
based on the application of the Semantic Web technologies
because of it is able to take advantage of the large number of
linked data that are available on the Web. The method of im-
plementation described as application case of the framework,
has tried to show that the system based on linked data is able
to start its operation with minimum user’s information.
Currently, the authors continue evaluating and validating
the framework under different scenarios. In addition, there are
been implemented some of the functionality of recommenda-
tion so that users can interact with the system
The incremental development of the system will determine
its strengths and limitations and will help to define the best way
to deploy an integrated platform of OERs recommendation.
ACKNOWLEDGMENT
This research has been partially funded by Regional
Government of Madrid (eMadrid S2013/ICE-2715) and the
scholarship provided by the Secretar´
ıa Nacional de Educaci´
on
Superior, Ciencia y Tecnolog´
ıa of Ecuador (SENESCYT).
REFERENCES
[1] F. Meyer, “Recommender systems in industrial contexts,” Ph.D.
dissertation, 2012. [Online]. Available: http://arxiv.org/abs/1203.4487v2
[2] T. Caswell, S. Henson, M. Jensen, and D. Wiley, “Open content
and open educational resources: Enabling universal education,”
The International Review Of Research In Open And Distance
Learning, vol. 9, no. 1, pp. 1–11, 2008. [Online]. Available:
http://www.irrodl.org/index.php/irrodl/article/view/469
[3] M. J. Pazzani and D. Billsus, “Content-based recommendation systems,”
Springer-Verlag, pp. 325–341, 2007.
[4] N. J. Belkin and W. B. Croft, “Information filtering and information
retrieval: Two sides of the same coin?” Communications ACM, vol. 35,
no. 12, pp. 29–38, 1992.
[5] M. Hiralall, “Recommender systems for e-shops,” Vrije Universiteit,
Tech. Rep., 2011.
[6] J. B. Schafer, D. Frankowski, J. Herlocker, and S. Sen, “The
adaptive web,” P. Brusilovsky, A. Kobsa, and W. Nejdl, Eds.
Berlin, Heidelberg: Springer-Verlag, 2007, ch. Collaborative Filtering
Recommender Systems, pp. 291–324. [Online]. Available: http:
//dl.acm.org/citation.cfm?id=1768197.1768208
[7] R. Burke, “Knowledge-based recommender systems,” in Encyclopedia
of Library and Information Systems. Marcel Dekker, 2000.
[8] B. Steichen, H. Ashman, and V. Wade, “A comparative survey
of personalised information retrieval and adaptive hypermedia
techniques.” Inf. Process. Manage., vol. 48, no. 4, pp. 698–
724, 2012. [Online]. Available: http://dblp.uni- trier.de/db/journals/ipm/
ipm48.html#SteichenAW12
[9] OCWC, “Ocw user feedback report,” Open CourseWare Consortium,
Tech. Rep., 2013. [Online]. Available: http://www.openedconsortium.
org/news/2013/06/ocwc-user-feedback- survey-results/
[10] T. I. Wang, K. H. Tsai, M. C. Lee, and T. K. Chiu, “Personalized learn-
ing objects recommendation based on the semantic- aware discovery
and the learner preference pattern,” Educational Technology & Society,
vol. 10, pp. 84–105, 2007.
[11] P. Winoto, T. Y. Tang, and G. McCalla, “Context in a paper
recommendation system with collaborative filtering,” The International
Review of Research in Open and Distance Learning, 2012.
[Online]. Available: http://www.irrodl.org/index.php/irrodl/article/view/
1243/2394
[12] M.-K. Khribi, M. Jemni, and O. Nasraoui, “Automatic recommendations
for e-learning personalization based on web usage mining techniques
and information retrieval,” International Forum of Educational Technol-
ogy & Society., vol. 12, no. 4, pp. 30–42, 2009.
[13] N. Manouselis, H. Drachsler, R. Vuorikari, H. Hummel, and R. Koper,
“Recommender systems in technology enhanced learning,” in Proceed-
ings of the 1st Workshop on Recommender Systems for Technology
Enhanced Learning (RecSysTEL), vol. 1, 2010, pp. 29–30.
[14] L. Chen, M. de Gemmis, A. Felfernig, P. Lops, F. Ricci, and
G. Semeraro, “Human decision making and recommender systems,”
ACM Transactions on Interactive Intelligent Systems, vol. 3, no. 3,
p. 17, 2013. [Online]. Available: http://dblp.uni- trier.de/db/journals/tiis/
tiis3.html#ChenGFLRS13
[15] J. Buder and C. Schwind, “Learning with personalized recommender
systems: A psychological view,” Computers in Human Behavior, pp.
207–216, 2012.
[16] M. Ruiz-Montiel and J. F. A. Montes, “Semantically enhanced
recommender systems.” in OTM Workshops, ser. Lecture Notes in
Computer Science, R. Meersman, P. Herrero, and T. S. Dillon,
Eds., vol. 5872. Springer, 2009, pp. 604–609. [Online]. Available:
http://dblp.uni-trier.de/db/conf/otm/otm2009-w.html#Ruiz-MontielM09
[17] J. Fink and A. Kobsa, “User modeling for personalized city tours,” Artif.
Intell. Rev., vol. 18, no. 1, pp. 33–74, 2002.
[18] V. A. A. Ayala, M. Przyjaciel-Zablocki, T. Hornung, A. Sch¨
atzle, and
G. Lausen, “Extending sparql for recommendations,” in Proceedings of
Semantic Web Information Management on Semantic Web Information
Management, ser. SWIM’14. New York, NY, USA: ACM, 2014,
pp. 1:1–1:8. [Online]. Available: http://doi.acm.org/10.1145/2630602.
2630604
[19] A. Passant, “dbrec music recommendations using dbpedia,” in SWC’10
Proceedings of the 9th international semantic web conference on The
semantic web, DERI. Springer-Verlag Berlin, 2011, pp. 209–224.
[20] C. Musto, P. Basile, P. Lops, M. de Gemmis, and G. Semeraro,
“Linked open data-enabled strategies for top-n recommendations,” in
Proceedings of the 1st Workshop on New Trends in Content-based
Recommender Systems co-located with the 8th ACM Conference on
Recommender Systems, Foster City, Silicon Valley, California, USA.,
2014, pp. 49–56. [Online]. Available: http://ceur- ws.org/Vol-1245/
cbrecsys2014-paper08.pdf
[21] T. Di Noia, R. Mirizzi, V. C. Ostuni, and D. Romito, “Exploiting the
web of data in model-based recommender systems,” in Proceedings of
the Sixth ACM Conference on Recommender Systems, ser. RecSys ’12.
New York, NY, USA: ACM, 2012, pp. 253–256. [Online]. Available:
http://doi.acm.org/10.1145/2365952.2366007
[22] C. Herzog, M. Luger, and M. Herzog, “Combining social and
semantic metadata for search in a document repository,” in Bridging
the Gep between Semantic Web and Web 2.0 (SemNet 2007), 2007,
pp. 14–21. [Online]. Available: http://www.kde.cs.uni-kassel.de/ws/
eswc2007/proc/CombiningSocialandSemantic.pdf
[23] J. Chicaiza, N. Piedra, J. L´
opez-Vargas, and E. T. Caro, “A
user profile definition in context of recommendation of open
educational resources. an approach based on linked open vocabularies,”
in 2015 Frontiers in Education Conference, FIE 2015, Texas,
United States, October 21-24, 2015, pp. 1–7. [Online]. Available:
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=7344314
[24] N. Piedra, E. Tovar, R. Colomo-Palacios, J. Lopez-
Vargas, and J. Chicaiza, “Consuming and producing
linked open data: The case of opencourseware,” Program,
vol. 48, no. 1, pp. 16–40, 2014. [Online]. Available:
http://www.scopus.com/inward/record.url?eid=2-s2.0-84893445335&
partnerID=40&md5=b6b2f262b3597d7dac38b5d731769018
[25] J. Chicaiza, N. Piedra, J. Lopez-Vargas, and E. Tovar-Caro, “Domain
categorization of open educational resources based on linked data,” in
Knowledge Engineering and the Semantic Web. Springer International
Publishing, 2014, pp. 15–28.