Personalized Recommendation Based Hashtags on E-learning Systems

Conference Paper (PDF Available) · November 2013with 157 Reads
DOI: 10.1109/ISKO-Maghreb.2013.6728109
Conference: 2013 3rd International Symposium ISKO-Maghreb
Cite this publication
The data generated by users on various social structures are growing exponentially over time. They become increasingly prodigious unmanageable and difficult to use. Therefore to easily find the content they produce among this mass of data, users label their own content using neologisms appointed hashtags. This practice attracts more and more the interest of researchers, because beyond the acquisition of knowledge, the Semantic Web approaches are also producing relevant information that may be used in practical situations. In this direction, we thought to exploit the activities of social Web users, mainly Hashtags. Hence, we focused on the identification of hashtags (as well as their different definitions) for personalized recommendation on e-learning systems. This paper aims at giving an insight on the pioneers' works and the opportunities raised by mixing the Social and the Semantic Web for education on one hand. And give the general architecture of our proposition and results obtained on the other hand.
Figures - uploaded by Frederique Laforest
Author content
All content in this area was uploaded by Frederique Laforest
Personalized Recommendation Based Hashtags on
E-learning Systems
eme Ghenname1,2, Mounia Abik1, Rachida Ajhoun1
LeRMA, Ecole Nationale Sup´
d’Informatique et Analyse des Syst`
e Mohammed V Souissi
Rabat, Morocco1
Julien Subercaze2,
Christophe Gravier2, Fr´
erique Laforest2
LT2C, T´
ecom Saint-Etienne,
e Jean Monnet, Universit´
e de Lyon
F-42000 Saint-Etienne, France2
Abstract—The data generated by users on various social
structures are growing exponentially over time. They become
increasingly prodigious unmanageable and difficult to use. There-
fore to easily find the content they produce among this mass of
data, users label their own content using neologisms appointed
hashtags. This practice attracts more and more the interest of
researchers, because beyond the acquisition of knowledge, the
Semantic Web approaches are also producing relevant informa-
tion that may be used in practical situations. In this direction,
we thought to exploit the activities of social Web users, mainly
Hashtags. Hence, we focused on the identification of hashtags (as
well as their different definitions) for personalized recomandation
on e-learning systems. This paper aims at giving an insight on the
pioneers works and the opportunities raised by mixing the Social
and the Semantic Web for education on one hand. And give the
general architecture of our proposition and results obtained on
the other hand.
Index Terms—Hashtags; Social Network; Semantic Web,
Learning process, personalized learning, recommendation of
contents, personalized learning environments.
Knowledge, education and learning are major concerns in
today’s society. Driven by new technologies, theories and
methods succeed, and the goal has always been to develop and
implement tools to make the learning process more effective
and relevant. However, effective and relevant means that the
process have to be in adequacy with times and societies.
The way we learn is motivated and guided by the society
where we live, and with the emergence of Web technologies,
learning move from normal classrooms to new environments,
it becomes, contextual, personal, and collaborative.
The Learning Management System (LMS) platforms focus
on publishing pedagogical materials to the learners, it also
allows to several students to be connected and share opinions
and knowledge in a synchronous or asynchronous way. Those
systems participate to increase the involvement of learners and
decrease their feelings of isolation. Actually LMS attempt to
adapt their content based on the learners profiles characteristic,
but implementing this important concept involves rethinking
the whole process of the learning experience to maximize
Many researchers support the improvement of learning
platforms, but far from being a simple extension, the question
raises original challenges in terms of technology as well as
educational. Thus maximizing the effectiveness of e-learning
experience means to consider several parameters, such as
profiles learners,their needs, and their focus in other words it
is necessary to determine which resources are required to the
learner profile. More precisely, we instantiate this problem on
discovering learners profiles without overwhelming end users
with questions or huge forms that are time-consuming. The
Social Web can be a potential solution to this problem, as they
are completely changing Web face by involving users in the
world of information and giving them the ability to participate
in the construction and dissemination of knowledge.
The social Web connects and captivates the attention of mil-
lion users from different ages, different cultures and different
languages. These web sites play a role in the establishment
of connections between people and their objects of interest.
As the number of users is constantly increasing, the needs of
new intuitive methods to represent, publish, and recommend
Web contents on these sites are becoming an utmost issue.
Since we are currently building the Semantic Web stack, its
formalism of data representation is therefore an interesting
way to describe social data. Those social data can be members
of a social network, the objects that arouse their attention,
their discussions threads, the links between the users and
their topics and resources of interest [1]. As a consequence,
a current intuition is that the Social Web can benefit from
the Semantic Web to ensure the structure and consistency of
its social data. This contributes to build the Web of (social)
Data, which aims to produce not only documents, but also a
set of interoperable and linked data on the Web. The primary
motivation is to make artificial agents that understand better
the social data so as to provide additional features to the Social
Web as we know it today.
The combination of the Social and Semantic Web allows
many improvements in the context of Web-based educational
systems. It can be of great help to enhance the learning expe-
rience. Indeed, it could drive us to provide custom learning
software to students, collect data related to the interaction
between students and the Web environment, discover the
services according to their needs, and make recommendations
of contents.
In this paper, we propose to review the opportunities iden-
tified from previous pioneers works for enhancing learning
when both the Social and the Semantic Web are involved
in the approach. We will then suggest research direction
that makes a socio-technical solution that allows linking the
2.0 applications, called collaborative and semantic dimension.
This will customize the useful information and facilitate the
access to the user (Semantic Web) and optimize the collective
learning situation by questioning the norms and forms of
electronic socialization serving the learning situation (Social
Recently, researchers and developers in learning technolo-
gies have started to combine Social Web and Semantic Web
techniques. Both paradigms aims at giving a well-defined
meaning for information, and opportunities such as learn-
ing individualization, free knowledge access, the opening of
training to new and wider audience despite the distance of
any kind (geographic, cultural, social, economic and trans-
actional). We can quote here the recent initiative of machine
Learning[2]. These devices have enabled the development of a
convivial relationship between trainers and trainees, and have
encouraged the learners independence and leads to the creation
of explicit and semantically-rich knowledge. Encouraged by
better conditions of access to various services, they are a real
source of educational revolution.
The experiences are increasing and many projects have been
initiated throughout the world. Researchers have therefore
thought to be involved more in the study of productivity
resulting from teaching and learning by exploiting the latest
technologies of the modern Web. Subsequently, various edu-
cational tools were made available.
The first attempts were about training systems as educa-
tional socials networks for learning languages, such as Campus
FLE Education, University of Leon, and Foreigners in Lille
or Exchanges Education Campus of the University of Lille
[3]. And many other commercials are available as lexxing,
Livemocha,, Babbel, etc. [4]. Those
tools are an excellent addition to the languages courses taken
at school, even an excellent substitute. The objectives of these
devices and communities of practice are: to promote authentic
communication, sub serve the linguistic and cultural exchanges
in the context of learning a foreign language, and finally to
develop comprehension and oral production for learners.
The blogs, thank to their mobile features for the provi-
sion of resources, and remote communication synchronous
and asynchronous, were also well positioned for educational
purposes. We mention some examples of experiences as
”EDUfranais2005” ownership by FLENET [5] for class note-
book blog. Or ”The Rapid E-Learning Blog”[6] which is a rich
collective learning environment. The idea of these proposals
was to use the blog as a support for a multi-channel suitability
for different learning styles of a group and as a toolbox to meet
individual needs.
Similarly we also have a variety of social platforms that fa-
cilitates the creation of groups of learners in learning commu-
nities, and enabling the sharing of ideas and artifacts from the
group. We find ”AlumnForce”[7] a social platform dedicated
to graduated networks that allows elders and organizations to
host their network and provide relevant jobs. ”Elgg”, a free
social platform adopted by the community of educators, is
structured on the concept of an extended personal profile, and
others like: Ning [8] and Personal Learning Environment [9]
etc. And more recently we have witnessed the development
of open source software used to configure virtual spaces
dedicated to training, where learners and instructors incarnated
through avatars can all over the world meet in the same
place and at the same time to share experience possess the
qualifications required to practice the interactions of group
phenomena which is necessary for knowledge construction.
For its part, the Semantic Web also came with a variety
of potentialities to contribute to the development of learning,
and we saw the first shades of platform-based Semantic Web
for teaching. The ”QBLS” system (Question Based Learning
System) [10] is a tool for resolving questions of supervised
work from semantically annotated course elements. It is a
way that serves not only to facilitate navigation, but also
encourage learners to seek knowledge actively with its ar-
chitecture mainly based on ontology. It follows a process
of questioning based on a teaching strategy called ”triple
rhythm” introduced in [11], and according to which the
training construction must comply with a division into three
stages: (1) raise the need of information in the heuristic phase,
(2) provide the information requested in the demonstration
phase; (3) process and assimilate the information. The purpose
behind this research initiative was to analyze the logs of the
students interactions with QBLS, and also assessing responses
to questions posed by the system to provide recommendations.
And if necessary, to correct the models of students training
activity for personalized pathways.
The platform ”TRIAL SOLUTION” [12] dedicated to the
publication of personalized documents, is based on existing
scientific books. Its general approach is to build a warehouse
of learning resources by extracting and annotating resources
with metadata from electronic text documents, to allow end
users semantic searches. The approach taken in this project
is similar to those of the Semantic Web that are designed
to automatically annotate learning resources from ontology.
Other similar approaches have been implemented in projects
such as ”Ariadne”[13] or ”Memorae” [14]. These projects also
proposed warehouses of learning objects. Resources are stored
and annotated with metadata standards for Ariadne and with
ontology for Memorae. No assumptions were made on the
resources origin, whether they were designed form warehouse
or they were extracted from existing documents.
The current social context points more and more new
technologies that appear in our daily lives. Education is not
setting aside of these developments, on the contrary it is at
the heart of this change. In this part we presented a review of
various existing tools dedicated to e-learning, whatsoever on
the side Social Web or Semantic Web. The list is not exhaustive
but brings us closer to contributions of each one of them in
terms of innovation and teaching practices renewal. We can
therefore conclude that each of these technologies represents
an asset for learning and the idea of their convergence could
be very beneficial for education. We will therefore explore the
possibilities of this convergence in the next section.
The major concern of today’s E-learning Systems is to
upgrade their skills and capabilities. Based on the principle
that the news paradigms of Web will offer rich seams of
diverse learning resources over and above the course materials
specified by course designers, a number of studies have
discussed the usefulness of analyzing social information by
using the semantic technologies. In other words, it makes sense
of information and its source, to give it structure for the best
possible use.
In [15], [16] and [17], the authors suggest how some current
developments in semantic technologies can be used to generate
personalized learning environments that will motivate learners.
They describe on one hand a Semantic Web-based e-learning
architecture and the importance of using metadata in the e-
learning field, and on the other hand, they list the challenges
which can be seen from the incorporation of semantic web in
a learning process and which may be essentially summarized
in: achieving interoperability between different educational
systems, automate the process of knowledge creation and
the structuring and the unification of educational data. To
open, share and reuse the content of education systems and
knowledge components.
Moreover, collaborative tagging has grown in recent years,
with sites that allow users to tag bookmarks, photographs and
other content. Consequently following the tagging practice
Folksonomy 1have emerged. Folksonomy is of a great interest
except that it lacks the semantic aspect related to the tags, since
users can use different labels to represent the same concept in
the annotation of a resource. In this vision, the researchers
realized the need to concoct the Semantic Web to Social Web,
and many of them have focused on approaches for the semantic
disambiguation of tags. In [18] the authors have related the
problems of lack of semantics that affect applications based on
Folksonomy to four main reasons: morphological variations,
the use of synonyms, the use of polysemic labels, and the
problem of using different labels to annotate a resource with
the level of users expertise in a particular field. To deal
with this problem, the authors proposed a context-based tag
disambiguation algorithm that selects the meaning of a tag
among a set of candidate DBpedia entries. They used a
common information retrieval similarity measure. Among the
contributions we also find [19] [20] taking as its starting point
the co-occurrences of a word recorded from a corpus and form
its different meanings by grouping its co-occurring according
to their similarity or dissimilarity.
More contributions have tried not only to use Semantic Web
in providing meanings and structure to data annotated and
facilitating their use, but they wanted to answer the question
”what can happen if we combine the best ideas from the Social
Web and Semantic Web?”. In this vision, new interventions
that have emerged regarding the semantic analysis of social
interactions for educational purposes have emerged. And so
[21] proposes a system for automatic generation of FOAF
profiles, which will be a searchable database in text or pictures,
allowing the identification of common characteristics among
members of TECFA (University of Geneva), and the location
of communities of interest and practice. [22] Applies tech-
niques of social network analysis to a database of annotations
extracted by a FOAF RDF crawl from profiles of LiveJour-
nal and build social networks from the properties ”knows”
and ”interest. We have also DBpedia a large-scale semantic
knowledge base, which structures socially created knowledge
on the Wikipedia, a wiki-based encyclopedia. DBpedia takes
advantage of the common patterns and templates used by
Wikipedia authors to gather structured information into a
knowledge base of socially created structured knowledge. So
with its capability of answering very specific queries, DBpedia
can serve as a very handy learning tool and is an excellent
example of the advantages that Social Semantic Web paradigm
brings to the educational domain.
Some research as [23] work has been done on leveraging
the above technologies in the e-learning domain. For example,
DERI Galway has developed a framework for extracting useful
knowledge published online in an informal way (e.g. wikis,
blog posts, forum posts), structuring the acquired knowledge
and putting it into use within LMSs. Or [24], which offers
an interesting and comprehensive approach for semi-automatic
generation of ontologies out of folksonomies In particular,
the authors suggest combining multiple online resources and
techniques such as processing techniques, statistical analysis
and social network analysis. Techniques that may be beneficial
to improve the semantic richness of the tags, and could also
be applied for analysis of tags students to identify students
sub-communities based on shared interests: annotated lessons
and/or tags used for annotation. Also [25] present an approach
for making explicit the semantics behind the tag space in
social tagging systems, so that this collaborative organization
can emerge in the form of groups of concepts and partial
ontologies, by using a combination of shallow pre-processing
strategies and statistical techniques together with knowledge
provided by ontologies available on the semantic Web.
Other important projects that utilize social and semantic in
order to provide support for interactivity at the Web scale
are Semantically-Interlinked Online Communities (SIOC)2an
initiative aimed at enabling the integration of user-generated
content and information contained in Web discussion methods.
The cornerstone of this initiative is the SIOC ontology3that
allows for machine readable and formal representation of all
data relevant in order to keep track of various kinds of Web dis-
cussions. Applied in educational settings, the SIOC ontology
enables gathering of data about all kinds of interactions that
a student has had on the Web, and allows for the inference of
additional knowledge about the student that can be beneficial
for improving his/her student model and his level of mastery
of some of the domain topics.
The Learning Object Context Ontologies framework
(LOCO) [26], aims at formally representing different kinds
of learning situations. It allows to formally representing all
particularities of the given learning context: the learning
activity, the learning content that was used or produced, and
the students involved. As a result, the framework integrates
a number of learning-related ontologies, enabling to formally
represent all the details of any given learning context, thereby
maintaining semantics in a machine interpretable format and
allowing for development of context-aware learning services.
We found also DEsign Patterns Teaching Help System
(DEPTHS)[27], it relies on the SocialSemantic Web paradigm
to prove learning systems and tools and provide students with
context-aware learning services. It’s developed using active
learning techniques, project-based learning, and collaborative
learning. The underlying philosophy of DEPTHS is based on
the fact that the major concern of todays software engineering
education is to provide students with necessary skills to
integrate theory and practice; to have them recognize the
importance of modeling and appreciate the value of a good
design; and to provide them with the ability to acquire specific
domain knowledge, in order to support software development
in specific domains.
As well through review of previous work, we can identify a
number of systems using the social Web and semantic Web, or
their combination in a learning environment. And also describe
the features and criteria used by each system from the two
facets of the modern Web. These features are summarized on
the Table I below.
In the next section we will show how we can exploit the
traces of users (Social Web) and give meaning (Semantic
Web). our approach will be the brick that binds the data
left behind by users on social networks with their profile as
learners on an e-learnig system (Figure 1).
A. Proposition
Despite many promising aspects given by the combination
of social and semantic Web as summarized in the Table I,
the semantic is still not widely adopted. This is mainly due
to difficulties in ontology creation and maintenance, and the
process of semantic annotation. Therefore many of current
perceptions confirm that the Social Web and the Semantic Web
dont oppose each other, but they can jointly cooperate to create
a social area of semantic technologies. The shared knowledge
on the Web leads to the creation of explicit and semantically-
rich knowledge representations. In other words the implication
of those technologies is substantial for the educational domain,
where students can immediately and efficiently get reply to
their request.
Our research works are within this perspective of integrat-
ing Social and Semantic Web in education, for better use
of pedagogy to improve teaching. For personalized learning
environments, the Semantic Web requires society-scale ap-
plications (e.g. advanced collaborative applications that use
shared data and annotations). So the social context is more
and more required, and social networks have become a new
technological context on the Web. A direction that currently
generates new forms of dialogue uses information about col-
lective intelligence, and their characteristics that may represent
real assets for training. Researchers have thus reached the
conclusion ”If we do not learn in a group, we will not learn
anything.” And so the idea is to take advantages of the growing
community, and integrate social work environments which has
become a priority.
Moreover, the paradigm of knowledge creation derived from
the Social Web can be effectively used to refine or update
generated ontologies according to Semantic Web standards
and best practices. And at the same time, the Social Web
can benefit from the paradigm of structured knowledge, repre-
sented with standard languages adopted in the Semantic Web
vision. The various studies cited in the previous section have
attempted to define in a better way the powers to operate
the traces that users leave on social structures, and organize
relevant information scattered on the Web with the aim of
creating data meaningful and easily usable by software agents,
in a process of automating complex tasks and assisting users
in their daily tasks.
Through the analysis of traces we can identify the behavior
and profiles of learners, and therefore to be informed of how
the students were active in social interactions, this will allow
us in one hand to help teachers to easily decide on how to
change their teaching approach to activate the students or make
them more focused on the relevant parts that will help them
in improving their level. And on the other hand, it helps to
make a categorization focus as by level of expertise or level of
prerequisites for the creation of virtual classroom learning. Our
approach focuses on the analysis of individual traces, and we
apply it also to analyze the interactions of the groups. Because
what we want to know, it does not restrict only on discovering
the learner profile but also its relationship with other people.
For a global vision of the learner and therefore build his
identity card containing its groups and specific guidance. But
despite all its advantages the social web is not suitable for
discussion too technical, it has a lot of noise, and suffers from
restrictions on information access due to the willingness of
students to keep some communications private.
Hence the great utility of an approach that converge the
usability of the social web in terms of use and creation of
information and capabilities of Semantic Web technologies
enable the creation of value through the structuring and
consistency of the data produced.
In our study, the communicative dimensions of Social Web
in real contexts are clarified, and the way that the practices
Social Web Semantic Web
Platforms Features
Features Databases Vocabularies
based on
logs of
annotation with
Campus FLE Education &
UStream TV
- - -
Exchanges Education Campus Facebook
marking and
- - -
DERI Galway
Forum, Wiki,
online on
structere in
an informal
knowledge -RDF
SNA social
Annotation based
on inexpensive
social sources of
based on
social book-
annotation with
Memorae 1! & Ariadne - -
based ontology
based Metadata
DBpedia Wikipedia
Data creation
based on
formal modeling -RDF
of constructivist pedagogies can be enhanced by the potential
of semantic technologies to support and enhance traditional
educational approaches is also. And then a future Intelligent
Learning Systems can be seen as collective knowledge sys-
tems, which are able to provide useful information based on
human contributions, and which improves as more people
participate for more and more personalized learning.
Our proposal is to increase the learning profile based on
hashtags enriched by the definitions. The goal is to create
an automated semantic profile based on traces of learners on
social networks with easier operating tags. In other words, we
want to present the learners profiles by their hashtags activities.
In the next section we discuss our approach on one side and on
the another one we propose an architecure (Figure 2) to enrich
the learning profile and make personalized recommendation.
B. Approach
Hashtags became a lightweight solution to classify and
search information on the Web 2.0. Unfortunately, is not a
piece of information by itself. The primary information is
the association (the tagging relation) that exists between a
hashtag and a resource. It is however of the utmost importance
to gain more knowledge on hashtags. Then to use them to
enrich the learners profiles, it is necessary to disambiguate
them in order to be used for purposes of recommendations on
e-learning environment. In this intuition we have developed
an approach for building a Folksionary. We coin the term
Folksionary as a porte- manteau term from terms folks and
dictionary. The Folksionary is a human readable document
that serializes the automatic generation of different synsets
for hashtags with several distinct meanings. More specifically
a Folksionary consists in a dictionary that for each hashtag
clusters its definitions in meanings, its a new kind of dictionary
to human users.
On our Work submitted to [28]. We present an approach that
provides a clustering of user-generated definitions for hashtags
into different senses, over any dataset of words along with their
user-generated definitions in natural English. The Figure 1
shows how our approach allows to link learners on LMS with
their activities on social networks.
The module named folksionary summarizes the different
steps for building a Folksionary. We perform a four-steps
process. First, we crawl hashtags definitions from online
services. Secondly, for each hashtag, we perform a pairwise
comparison of its definitions by computing a distance between
pairs of definitions. At the third step, we apply a clustering
algorithm for each hashtag in order to group its definitions
into similar meaning clusters. Lastly, we export these results
under the form of a human-readable document with a look
very close to a standard dictionary.
1) Step1:Crawl hashtags definitions:In this step different
sources of data on the Web contain user written definitions
of hashtags in natural language. For instance, or are well-known online hashtags dictionaries. In
this first step, we crawl some hashtags and their definitions
from such sources. The scrapping process extracts definitions
from each given page and, using a language classifier, only
english definitions are kept in our database.
2) Step2:Distance between hashtags:The user-generated
definitions for a given hashtag can be redundant, i.e. some
definitions can describe the same meaning. Our goal in this
step is to measure the semantic-relatedness between definitions
in order to provide an input for the clustering phase. We
use the Extended Lesk algorithm because among different
techniques involving an external knowledge base, the Extended
Lesk algorithm has proven to be one of the most efficient[29].
3) Step3:Clustering of definitions:In the previous step
we generate a graph providing these relationships. This graph
is used to cluster hashtags based on their meanings. To
achieve the clustering we use the Markov Clustering algorithm
(MCL)[30]. A comparative analysis have shown that MCL
is remarkably more robust than other clustering techniques
[31]. It produces good clustering results mainly because the
algorithm scales well with increasing graph size, it is robust
against noise in graph data even if it cannot find overlapping
clusters. Also we are not constrained to specify a number of
clusters beforehand. This is suitable for our approachas we
have no a priori information regarding the number of clusters.
4) Step4:Formatting a folksionary:One of the objectives
of a Folksionary is to provide a new kind of dictionary to
human users. Therefore we output the in-memory model of
the folksionary in a format close to a traditional dictionary.
This output is a PDF file that organizes hashtags entries in
an alphabetic order. Each hashtags is presented with all its
meanings, and we list in each meaning all the definitions that
were clustered. The folksionary PDF file containing all tags
is available online at
5) Step5:Evaluation:In order to complete the quantitative
analysis of our folksionary, a qualitative analysis is needed.
It consists in measuring the distance between the generated
folksionary and an expected Ground Truth. Given that to our
knowledge no other folksionary exists in the literature, we
were brought to build this Ground Truth manually for all
polysemic tags. We have built an ad hoc Web application, and
participants have manually built the ground truth by clustering
hashtags’ definitions into meanings. We have conducted obser-
vations on the entire dataset in order to measure the distance
between the Ground Truth partitioning and the automatic par-
titioning generated by our approach. For this purpose we have
performed a straightforward evaluation using a metric adapted
to our dataset Average Conditional Probability (ACP)[32].
6) Step6:Next close step:In the continuation of work,
we will bridge the gap between what was already done to
this stage and learning profile defined according to a given
standard. We are primarily interested at enriching the fields
interest as the hashtags are suitable to provide interesting
information on the interests of users of social networks.
In this work we propose our architecture that will link the
traces of users on social networks (hashtags) with their e-
learning profile Figure 2. The different layers of our architec-
ture are: Layer1: Construction of Folksionary for learning an
Fig. 1. Process for enrichement of Learner Profile
ontology of tags on one side and to infer the user personomy
on the another. Layer2: Feeding the different fields of learners
profile(profile defined according to an e-learning standard).
Layer3: Recommendation engine based on the user profile
and the application domain. Layer4: Application layer that
allows to implement the recommendation process through
personalized prediction of items attracting the interest of the
learner in the LMS.
We are currently at the step of generating an ontology
from our Folksionary, we are currently at Step generating
an ontology from our first part folksionary. The remaining
steps will be our future work, more details will be given
in a forthcoming paper in which we will implement our
architecture and we will apply it in a use case.
C. Results
In this section we summarized our results, and our perspec-
tives for the future works. We have introduced the concept
of folksionary which consists in a dictionary that for each
hashtag clusters its definitions in meanings. We have also
defined a four-steps process to build a folksionary. First we
gather all definitions by crawling online services, we then
apply a semantic distance measure between definitions for
each hashtag. We perform a clustering that groups similar
definitions into distinct meanings clusters. Clusters are finally
presented under the form of a human-readable folksionary.
We have conducted a validation of this process: we have
developed a web application to build the Ground Truth with
a human end-user who manually clusters the definitions. A
pairwise evaluation of the results of our clustering process in
comparison with the Ground Truth has been conducted. The
Evaluation results shows that our approach works not only in
theory but also in practice: It’s performing well on most of
the datasets and producing good results for definition sense
clustering, by approaching Ground Truth with 89.21831%.
The very close next step concerns the choice of a standard
for modeling learner. Generating an ontolgie from the semantic
relationships between the hashtags on the Folksionary. And
then establish the link between the ontology is the modeling
standard in order to to enhance the Learners profiles for
personalized recommendation on e-learning systems.
This work explores the potential of social and semantic Web
technologies for e-learning systems and recommending per-
sonalized contents. In this paper, we have focused our attention
on the proposal of guidelines for tracking the different types of
contributions, activities and learners conversations. And also
make usable their spontaneous collaborative classifications on
social networks. We propose a methodology for exploiting
hashtags in the automatic generation of learnres semantic pro-
files. This work allows to link what was done previously in our
work concerning the automatic approach for structuration of
hashtags definitions into synonym rings. We present the output
as a so-called Folksionary. The very close next step concerns
the generation of ontology from our Folksionary. We then plan
to develop techniques to discover other semantic relationships
between user hashtags : synonymy, hyperonymy, or part-of.
in order to expand the learner profile, as well as the learning
preferences and then make personalized Recommendation of
pedagogical content.
[1] Approaches to social research (2nd ed.).Singleton Jr., Royce A.; Straits,
Bruce C.; Straits, Margaret Miller New York, NY, US: Oxford University
Press. (1993). xx 572 pp.
[2] Andrew Ng Stanford Artificial Intelligence Lab,
[3] Franoise Wolf Mandroux, ”L’apprentissage des langues distance de
l’ingnierie la pdagogie”, centre de ressources et d’ingnierie documentaire,
Avril (2011).
[4] LANGUES MAG Rseau dapprentissage des langues
[5] Mario TOM, WEBPRATIQUE - FLE. Apprendre
le franais, dcouvrir Internet, crer des pages web
[6] The Rapid E-Learning Blog,
[7] AlumnForce Associations des Anciens lves Rseaux Campus et Profes-
sionnels coles & Universits ,
Fig. 2. Architecture for Personalized Recommendation Based Hashtags on E-learning Systems
[8] Ning Tutorial,
[9] PLEF Personal Learning Environment Framework,
[10] S.Dehors, C. Faron-Zucker, R. Dieng-Kuntz, QBLS:Semantic Web Tech-
nology for E-learning in Practice , In 15th International Conference on
Knowledge Engineering and Knowledge Management, Czech Republic,
2nd-6th October 2006.
[11] Blanc, D. (2001), La bote outils du formateur, Editions dorganisation
Eyrolles, troisime dition.
[12] W. Lenski, E. Wette-Roch. The Trial-Solution Approach to Document
Re-use, in Proc. of Workshop on Electronic Media in Mathematics,
Coimbra, Portugal, 2001.
[13] E. Duval, E. Forte, K. Cardinaels, B. Verhoeven, R. Van Durm, K. Hen-
drikx, M. Wentland-Forte, N. Ebel, M. Macowicz, K. Warkentyne, and
F. Haenni, The ARIADNE Knowledge Pool System, in Communications
of the ACM 44 (2001), no. 5, pp. 73-78.
[14] Abel M.-H., Lenne D., Ciss O. (2002) e-learning et Web smantique :
le projet MEMORAe In actes lectroniques des journes scientifiques web
smantique, Ivry sur Seine, 10-11 octobre.
[15] M.Nilsson, Matthias Palmr Semantic Web Meta-data for e-Learning-
Some Architectural Guidelines Proceedings of the 11th World Wide Web
Conference (2002)
[16] Aroyo, Lora and Dicheva, Darina (2004): The New Challenges for E-
learning: The Educational Semantic Web. In Educational Technology &
Society, 7 (4) pp. 59-69.
[17] Anderson, T., & Whitelock, D.(2004). The Educational Semantic Web:
Visioning and Practicing the Future of
[18] Andrs Garca-Silva, Martin Szomszor, Preliminary Results in Tag
Disambiguation using DBpedia K-CAP09, September 1-4, 2009, Redondo
Beach, California, USA. Copyright 2009 ACM.
[19] Beate Dorow et Dominic Widdows, Discovering corpus-specific word-
senses , 10th Conference of the European Chapter of the Association for
Computational Linguistics (Conference Companion) pp. 79-82, (2003)
[20] Ferret, Desambigusation automatique partir d’espaces vectoriels, mul-
tiples cluteerises, http ://, (2004)
[21] Melie Genet, Favoriser la perception des communauts en ligne et la
diffusion des connaissances en rsumant les informations publiees sur des
pages personnelles, http ://
[22] J C Paolillo and E Wright, Social Network Analysis on the Semantic
Web Techniques and Challenges for Visualizing FOAF in Book Visual-
izing the semantic WebXmlbased Internet And Information, (2006).
[23] A Passant et P Laublet, Meaning Of A Tag A Collaborative Approach to
Bridge the Gap Between Tagging, and Linked Data LDOW2008, (2008)
[24] Van Damme, C., Hepp, M., & Siorpaes, K. (2007). FolksOntology: An
integrated approach for turning folksonomies into ontologies. Proceedings
of the Workshop Bridging the Gap between Semantic Web and Web 2.0
at the 4th European Sematic Web Conference (pp. 5770).
[25] Lucia Specia and Enrico Motta Integrating Folksonomies with the
Semantic Web The Semantic Web: Research and Applications Lecture
Notes in Computer Science, 2007, Volume 4519/2007
[26] Jovanovic J., Knight C., Gasevic D., Richards G. (2007). Ontologies
for Effective Use of Context in e-Learning Settings. IEEE Educational
Technology & Society, 10(3), 47-59.
[27] Jeremic, Z., Jovanovic, J., Gasevic, D. (2009). Semantic Web Technolo-
gies for the Integration of Learning Tools and Context-aware Educational
Services. In: Proceedings of the 8th International Semantic Web Confer-
ence, Washington, DC, USA (Lecture Notes in Computer Science, Vol.
5823), pp. 860-875.
[28] M. Ghenname, J. Subercaze, C. Gravier, F. Laforest, M. Abik, R.
Ajhoun.Learning hashtags polysemy from crowd-sourced definitions.The
7th ACM WSDM Conference Web Services Distributed Managemen
(WSDM14) New York City, February 24-28, 2014.
[29] S. Banerjee and T. Pedersen. An adapted lesk algorithm for word sense
disambiguation using wordnet. In CICLing’02, pages 136-145, 2002.
[30] A. J. Enright, S. Van Dongen, and C. A. Ouzounis. An efficient algorithm
for large-scale detection of protein families. Nucleic Acids Research,
30(7):1575-1584, Apr. 2002.
[31] S. Brohee and J. van Helden. Evaluation of clustering algorithms for
protein-protein interaction networks. BBMC Bioinformatics, 7:488, 2006.
[32] G. R. Burset M. Evaluation of gene structure prediction programs, 1996.
  • Article
    Full-text available
    E-learning adaptation has become the most important method that facilitates access to the appropriate content. Adaptive approaches consist of reducing the problems of incompatibilities between learner's cognitive abilities and educational content's difficulties. In some cases, the adapted curriculum cannot meet learner's skills completely seen its incoherent structure, its unsuitable methodologies and sometimes its complexity. Therefore, we need to measure the convenience of the content material to improve it and ensure learners' satisfaction. In other words, it is necessary to estimate its appropriateness to each learner. That is why; we have proceeded by using decision tree (DT) algorithm which is a supervised data mining method. It helps to predict the convenience of the proposed content material for learners. Our system consists of classifying learning material into two classes: "good" if it is convenient, and "anomaly" if not. To achieve that, we have used an intelligent agent called Classifier Agent (CLA). It tracks learner's behavior by collecting a set of attributes like score, learning time, and number of attempts, feedback and interactions with the tutor. Then, he calculates the predictive attribute by using the (DT) algorithm. The finding algorithm shows that the score is the most crucial indicator gives us more information about the conformity of curriculum to learners, followed by learning time, feedback and number of attempts. © International Journal on Advanced Science, Engineering and Information Technology.
  • Chapter
    This paper presents a real case of tracking conversations and participation in social networks like Twitter and Google+ from students enrolled in a MOOC course. This real case presented is related to a MOOC course developed between January 12 and February 8, 2015, in the iMOOC platform, created as result of the collaboration by Technical University of Madrid, University of Zaragoza and University of Salamanca. The course had more than 400 students and more than 700 interactions (publications, replies, likes, reshares, etc.) retrieved from the social both social networks (about 200 interactions in Twitter and 500 in Google+). This tracking process of students’ conversations and students’ participation in the social networks allows the MOOC managers and teachers to understand the students’ knowledge sharing and knowledge acquisition within the social networks, allowing them to unlock the possibility of use this knowledge in order to enhance the MOOC contents and results, or even close the loop between the students’ participation in a MOOC course and the parallel students' usage of social networks to learn, by the combination of both tools using adaptive layers (and other layers like the cooperation or gamification like in the iMOOC platform) in the eLearning platforms, that could lead the students to achieve better results in the Learning process.
  • Chapter
    With the accelerated pace of life, e-book reading has become a way for people to learn and play in their leisure time. When faced with a large number of e-books, the recommendation system and e-book rankings have become the entrance for readers to quickly find books of their own interest. Through the current situation of cash development of e-books, this paper analyzes the abnormality of information recommended by personalities, and discusses the problems of intelligent detection systems.
  • Conference Paper
    Many educational institutes are recording classroom lessons to improve the students’ learning process. Unfortunately, a simple recording does not necessarily imply benefits. Indeed, when pro- ducing video lectures, there are different challenges that must be faced: production costs, accessibility, usability, video indexing, just to name a few. In this paper, we share our experience building ONELab, a system designed to capture, record, edit and stream video lectures. ONELab was designed to be scalable and to have low-cost implementation and maintenance. The system has been used in the 2017-18 Academic Year to manage the 49 courses offered by the five degrees available at our Department. In numbers, it sup- ported 1,054 freshman students and produced 1,376 video lectures (for a total of 2,064 hours). The usage analysis showed that stu- dents appreciated the system and a comparative analysis between students who used the system and students who did not use the system, showed that the former passed more exams (+97.8%), had better grades (+8%) and acquired more credits (+105%).
  • Article
    Two major educational strengths that MOOCs provide are informal learning and harnessing the collective intelligence of the students and the interactions among other users like former students, future students, business professionals, other universities, etc. These features may lead to the emergence of new sustainable in time educational elements wherein knowledge and learning continue enriching once the course finished. At present, one of the main limitations of the MOOC platforms is the lack of social open tools to enhance and take advantage of the collective intelligence generated in the course. This article proposes a new model to allocate informal learning and collective intelligence in MOOCs using external virtual learning communities through social networks, based on Google +. The main aim of this article is to assess the virtual learning community performance and analyze the interactions and the kinds of learning that take place inside the community and over time. A case of study of a MOOC course with Google + community is presented.
  • Article
    Full-text available
    : I (Terry) first became interested in the semantic web from reading Berners-Lee's original works and following first generation developments of semantic web technologies in information science, e-business and health fields. I then began including the ideas in talks I gave at various conferences and forums in 2003. Naturally, I became curious about what other educators were doing with the semantic web and so Googled the term, "education semantic web". Much to my surprise and disappointment, I found that most of the references were to my own admittedly introductory and visionary comments made in these speeches. Where was the real work, innovation and actual prototype development? Fortunately, we were able to locate this type of work and we believe that most of the leading researchers in the area of the educational semantic web have contributed to this special issue. Of course, if we have missed your work, we welcome comments and URLs in the discussion areas of the special issue. Editors: Terry Anderson and Denise Whitelock.
  • Article
    The Semantic Web promises to provide new applications for Internet users through the use of RDF metadata attached to various information resources on the Web. Yet it is somewhat unclear who will provide the metadata, or what will motivate people to provide it, let alone the exact nature of the applications the Semantic Web will ultimately support. What will the "killer app" of the Semantic Web be, and what shape will it take? An answer to this question may have already arisen, in the form of the Friend-of-a-Friend (FOAF) vocabulary. The FOAF project was begun in 1999 to explore the application of Semantic Web technologies (RDF/XML) to describing people's personal details: their professional and personal lives, their friends, interests, and other social dispositions. Its main product is the FOAF vocabulary, an RDF/XML namespacewith elements defined for describing an individual's social sphere (Brickley and Miller, 2003). Recently the FOAF vocabulary has been adopted by many large Web-logging ("blogging") and social networking software sites, such as LiveDoor, LiveJournal, and others.Weblogs, a recent Internet phenomenon, are diary-like sites usually consisting of entries in reverse-chronological order.Herring, et al. (2004) situateWeblogs as a genre bridging between extant media technologies and new forms of computermediated communication. The contribution ofWeblogs to the SemanticWeb comes from the design of the supporting software to automatically generate RDF/XML files including RSS feeds, and now FOAF. The popularity ofWeblog hosting sites, and their ability to automatically generate RDF, has had a large impact on the Semantic Web. Swoogle (, at present the largest fully automatic SemanticWeb document aggregator, currently lists 19 large Web-logging sites in its top-50 index of sites with Semantic Web content (LiveDoor is top-ranked with 9,473 documents in Swoogle and LiveJournal second with 7,690), and collectively blogging sites are responsible for 45% of the Semantic Web documents collected by Swoogle. Even from these sites, Swoogle crawls only a small fraction of the available FOAF documents: LiveJournal automatically generates FOAF files for each of its 6.7 million users, several times more than the number of documents in Swoogle.
  • Article
    Full-text available
    Meta-data is the fundamental building block of the Semantic Web. However, the meta-data concept is too loosely defined to provide architectural guidelines for its use. This paper analyzes important uses of meta-data in the e-learning domain, from a pedagogical and philosophical point of view, and abstracts from them a set of fundamental architectural requirements for Semantic Web meta-data. It also describes some flexible generic techniques for working with meta-data, following these requirements. Finally, the paper describes a Semantic Web-based e-learning architecture based in these requirements and techniques currently under development at the Knowledge Management Research Group at CID (Centre for user oriented IT Design) at KTH, the Royal Institute of Technology in Stockholm. This architecture builds on Edutella, a peer-to-peer meta-data exchange network, and a technique called conceptual modeling using the Conzilla concept browser, a new kind of knowledge management tool for conceptual navigation and exploration. The
  • Article
    Full-text available
    This paper introduces MOAT, a lightweight Semantic Web framework that provides a collaborative way to let Web 2.0 content producers give meanings to their tags in a machine-readable way. To achieve this goal, this approach relies on Linked Data principles, using URIs from existing resources to define these meanings. That way, users can create inter-linked RDF data and let their content enter the Semantic Web, while solving some limits of free-tagging at the same time.
  • Conference Paper
    One of the main software engineers’ competencies, solving software problems, is most effectively acquired through an active examination of learning resources and work on real-world examples in small development teams. This obviously indicates a need for an integration of several existing learning tools and systems in a common collaborative learning environment, as well as advanced educational services that provide students with right in time advice about learning resources and possible collaboration partners. In this paper, we present how we developed and applied a common ontological foundation for the integration of different existing learning tools and systems in a common learning environment called DEPTHS (Design Patterns Teaching Help System). In addition, we present a set of educational services that leverages semantic rich representation of learning resources and students’ interaction data to recommend resource relevant for students’ current learning context.
  • Conference Paper
    Full-text available
    While tags in collaborative tagging systems serve primarily an indexing purpose, facilitating search and navigation of resources, the use of the same tags by more than one individual can yield a collective classification schema. We present an approach for making explicit the semantics behind the tag space in social tagging systems, so that this collaborative organization can emerge in the form of groups of concepts and partial ontologies. This is achieved by using a combination of shallow pre-processing strategies and statistical techniques together with knowledge provided by ontologies available on the semantic web. Preliminary results on the and Flickr tag sets show that the approach is very promising: it generates clusters with highly related tags corresponding to concepts in ontologies and meaningful relationships among subsets of these tags can be identified.
  • Conference Paper
    Full-text available
    This paper presents an adaptation of Lesk’s dictionarybased word sense disambiguation algorithm. Rather than using a standard dictionary as the source of glosses for our approach, the lexical database WordNet is employed. This provides a rich hierarchy of semantic relations that our algorithm can exploit. This method is evaluated using the English lexical sample data from the Senseval-2 word sense disambiguation exercise, and attains an overall accuracy of 32%. This represents a significant improvement over the 16% and 23% accuracy attained by variations of the Lesk algorithm used as benchmarks during the SENSEVAL-2 comparative exercise among word sense disambiguation systems.