Classification and Indexing of Web Content
Based on a Model of Semantic Social Bookmarking
Antonello Angius
1
, Giulio Concas
2
, Dino Manca
3
, Filippo Eros Pani
2
and Georgia Sanna
3
1
Centro Regionale di Programmazione, Regione Autonoma della Sardegna, via Battisti, Cagliari, Italy
2
Department of Electrics and Electronics Engineering, University of Cagliari, Piazza d'Armi, Cagliari, Italy
3
Experteam srl, Via Zara 11, Cagliari, Italy
Keywords: Multimedia Content, Social Networks, Social Bookmarking, Taxonomy, Folksonomy.
Abstract: One of the key challenges in Information Technology is finding a way to organize the knowledge present on
the Web. This led to years of research on the integration of information, on the Semantic Web and related
technologies. Information Search and Retrieval from the Web occur through a process of content
disambiguation and search engines use algorithms and software agents in order to meet the needs of users
and advertising buyers. Ex-post analytical agents and tools are becoming more pervasive, so much to cause
increasing problems of privacy. Our work proposes an innovative approach of content disambiguation that
overturns the ex-post semantic analysis of contents, because it deals with an ex-ante classification
conducted on two axes: vertical one (hierarchical and taxonomic axis) and horizontal one (folksonomic axis
through tags or keywords). This method, which is based on the logic of social bookmarking and focuses on
semantic tagging, represents a new frontier in information architecture because it introduces a new way of
classification made by people using keywords that have a specific lexical and semantic value. This approach
will allow people to create a knowledge base of Web contents characterized by a precise semantic
definition.
1 INTRODUCTION
Search engines use several software agents and
original algorithms to meet needs of private users
and advertising buyers. However, in order to achieve
satisfying results in searches through “words”,
information on the Web should be semantically
connoted.
On the user's side, it is often difficult to filter
properly the information through thousands of
search results, while on the advertising's side there is
a relevant percentage of "contextual" messages that
are outside the target and therefore they are ignored
or unwelcome by people.
The approach proposed in this work is based on a
technique of disambiguation that overturns the ex-
post semantic analysis of contents, because it deals
with an ex-ante classification conducted on two
axes: vertical one (hierarchical and taxonomic axis)
and horizontal one (folksonomic axis through tags or
keywords).
The basic idea of this approach is to allow
people to classify contents with an innovative
approach related to social bookmarking: users will
be enabled to bookmark a content, classify it
semantically and add it to the knowledge base.
According to the logic of social bookmarking,
the knowledge base will be enriched/supplemented
by other users with further classified contents. This
method will promote the development of a new
collaborative model of Web 2.0, where people create
the knowledge base and classify interesting contents
drawn from social networks, other UGC (User-
Generated Content) or Web sites (Lunesu et al.,
2011) (Lunesu et al., 2011).
The paper is structured as follows: Section Two
shows an overview of the knowledge bases used to
classify contents. In Section Three we present the
most relevant components of the system (plugin,
classification model, methods of interaction among
users). Section Four includes the conclusions and
some reasonings about our work. Section Five deals
with some possible future developments.
313
Angius A., Concas G., Manca D., Eros Pani F. and Sanna G..
Classification and Indexing of Web Content Based on a Model of Semantic Social Bookmarking.
DOI: 10.5220/0005139303130318
In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2014), pages 313-318
ISBN: 978-989-758-050-5
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
2 KNOWLEDGE BASES USED
FOR CLASSIFICATION
Starting point of this work is the examination of
baseline studies and tools, including UGC, which are
used to disambiguate and classify word meanings, in
order to categorize contents with disambiguated
terms (Passant and Laublet, 2008) (Suchanek et al.,
2008) (Zhong, 2008) (Agirre and Soroa, 2009)
(Murgia et al., 2010).
Consequently, in this section we describe the
major scientific lexical-semantic projects such as
WordNet and BabelNet, which assurance a correct
disambiguation of terms at the lexical and semantic
level, and resources which are collaboratively run by
largely volunteers according to the logic of Web 2.0
(Berners-Lee et al., 2001), such as the well-known
Wikipedia and DBpedia project.
Among the other interesting projects on tags
meaning disambiguation and RDF metadata
processing, we mention PiggyBank, which is a tool
integrated into the Web browser that allows users to
extract information items from within Web pages
and save them in Semantic Web format, replete with
metadata (Huynh et al., 2005) and DogmaBank,
which is a social bookmarking tool that can be used
to tag Web pages with freely chosen keywords or
with concepts. (Spyns et al., 2006).
2.1 WordNet
The proposed technique is based on WordNet, a
large semantic-lexical database of English, whose
development began in 1985 at the Cognitive Science
Laboratory of Princeton University under the
direction of linguists and psychologists.
WordNet aimed at developing a more intuitive
dictionary/thesaurus and a support for Artificial
Intelligence applications and for technologies related
to automatic text analysis. It groups nouns, verbs,
adjectives and adverbs into sets of cognitive
synonyms (synsets) by means of semantic relations
(between word meanings) and lexical relations
(between word forms). Its database includes 117000
synsets, which are interlinked by means of a small
number of conceptual relations, contains a brief
definition and, in most cases, one or more short
sentences explaining the use of synset members.
(Miller, 1995).
WordNet differs from a common thesaurus
firstly because it creates relationships not only
between word forms (strings of letters) but also
between the specific sense of words. Secondly, it
labels the types of lexical or semantic relationship
among words, whereas a common thesaurus groups
words just on the level of meaning similarity
(Fellbaum, 1998).
2.2 Wikipedia
Wikipedia, the well-known universal "free
encyclopedia that anyone can edit”, well represents
the spirits in which free software is developed: its
keywords could be collaboration, sharing and
improvement.
It is run largely by volunteers and it is not
regulated by a central committee. However, there are
review processes: although everybody can
contribute, regardless of the education or expertise
level, the user is warned about the possibility of
change/deletion/redistribution of contents, where the
last possibility depends on the Creative Commons
Licence.
Wikipedia articles, represented as pages
(WikiPages), contain information on a concept (such
as Play-Theatre) or on an entity (like William
Shakespeare). Each WikiPage's title includes the
concept's term (Play) in addition to an optional label
enclosed in parentheses that, in cases of ambiguity,
specify its meaning (Wu and Weld, 2007) (Adar et
al., 2009) (Toral et al., 2009).
2.3 DBpedia
DBpedia project focuses on converting Wikipedia
information into structured knowledge, which is then
made available on the World Wide Web as RDF
(Resource Description Framework), and allows to
execute complex queries on Wikipedia contents and
to link the extracted data to other datasets (Lehmann
et al., 2014) (Morsey et al., 2012). It presents the
following features:
1) a framework for data mining, which converts
Wikipedia's content into RDF (Brickley et al., 1999);
2) Wikipedia's content is available through a huge
and multi-domain RDF dataset;
3) the dataset can be connected to other datasets with
a resulting Web of data that contains, in all, 2 billion
RDF triples (Garcia et al., 2009).
2.4 BabelNet
BabelNet is a semantic network in constant
evolution, which includes encyclopedic and
lexicographic items. It has been developed by
Roberto Navigli, Paolo Ponzetto (and others) at the
University of Rome “La Sapienza” within the “ERC
Starting Grant MultiJEDI MultiJEDI” (Navigli and
Ponzetto, 2010).
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
314
Main aim of BabelNet is to provide a unified
resource which integrates Wikipedia and WordNet
through a system of automatic mapping: WordNet
furnishes the missing relationships in Wikipedia and
Wikipedia adds the multilingual dimension.
Multilingualism is another purpose of BabelNet:
the lexicalized concepts belonging to the different
languages are collected from Wikipedia through
interlingual links and, in case of poor languages,
lexical gaps are filled with the aid of automatic
translation systems.
Similarly to what has been proposed by
WordNet, in Babelnet words are grouped into sets of
synonyms called Babel synsets. Each Babel synset
provides a given meaning and contains all the
synonyms which express that meaning in several
languages (Navigli, Ponzetto, 2012).
BabelNet 2.5 (May 2014) is obtained from the
automatic integration of six linguistic resources:
WordNet 3.0, Wikipedia, Open Multilingual
WordNet, OmegaWiki, Wiktionary, Wikidata.
3 THE PROPOSED APPROACH
The proposed approach is based on an ex-ante
classification conducted on two axes: vertical one
(hierarchical and taxonomic axis) and horizontal one
(folksonomic axis through tags or keywords). This
technique aims at being used by people who want to
classify semantically information items and add
them to the bookmark. All the contents classified by
individual users will be categorized in a large
knowledge base, that could be shared with other
users.
According to the logic of social bookmarking
and to a new collaborative model of Web 2.0, other
users will enrich and integrate this knowledge base
by classifying other contents, which could also be
shared and commented on social networks, forums
and blogs (Concas et al. 2008). People will be able
to use this tool to find contents, searching them by
means of their classification or their disambiguated
tags.
The set of pre-classification services and results
matching is run by a personal classification engine.
This approach involves a main axis of innovation
which consists of a new model of information
classification and which includes two integrated
types of skills related to two different fields: on one
side the information architecture and on the other
side the software.
On the side of information architecture, we are
developing a new structure of thesaurus (a linguistic
database that contains logical, semantic and
hierarchical relations among terms) based on
BabelNet.
This tool aims to be easily used and integrated
by people to classify their contents of interest on a
horizontal axis (in which tags refers to the
folksonomy based on the thesauri) and on a vertical
axis (regarding the hierarchical-taxonomic assigning
of tags), with a resulting semantic disambiguation of
the classified contents.
3.1 Plugin Software
From a software point of view, the system allows
users to select through a specific plugin a content on
a social network (such as Facebook) or on any Web
site. The main CMS environments (e.g. Drupal,
Joomla!, Wordpress,) for Social Networks
(Facebook, Twitter, etc.) and forums are under
development.
In addition, we have developed a bookmarking
toolbar that enables users to select areas of text and
images which can be classified and added to the
bookmark, with the purpose of making them
available on the knowledge base.
New features of social semantic bookmarking
will be provided by this plugin, with the added value
of a total ex-ante disambiguation, whose functional
and social implications are quite extensive. It is
therefore evident that our work does not provide a
“simple” way of bookmarking contents which
integrates or it is alternative to classic old social
bookmarking systems such as Del.icio.us, which is
a social bookmarking Web service for saving,
organizing and discovering links on the Web, that
can be used to tag information items with freely
chosen terms. disambiguation of the classified
contents.
3.2 Relationships among Users
The system enables to develop a knowledge base
which will be different from those provided by
common search engines because it will be based on
an ex-ante classification, socially controlled, and
therefore more precise than those provided by ex-
post classification. It will also allow new ways of
relationships among users, supporting the
communication and the exchange of information
through new forms of interaction based on common
thematic interests.
This new technology will be completely
different from the traditional “friendship on Web”
model, called FOAF (Friend Of A Friend), which is
ClassificationandIndexingofWebContentBasedonaModelofSemanticSocialBookmarking
315
normally developed through concatenation
processes. People will receive notifications
whenever a content pertaining to their area of
interest is published, supporting therefore the
interaction with those people who have inserted the
content into the knowledge base.
This new form of interaction, depending on
thematic basis, is also different from that of
"groups", which is typical of social networks and
some forums, because in our system the first contact
among users does not depend on a general adherence
to themes, titles or designations of groups.
The technique of disambiguation is activated by
the personal interests classified by the user through a
built-in feature, in order to automatically perform
searches and daily crossings with all the informative
items having the same or similar classification.
This means that once people set a "dashboard" of
contents and personal interests (freely configurable
and upgradeable), they will be able to view at each
access all the contents and information inserted by
other users which are consistent to their dashboard.
Consequently, people will no longer have to search
information and contacts.
3.3 The Classification Model
On the side of information architecture the
classification model, which aims at the pre-indexing
of content made by users, is developed on the basis
of the latest scientific and empirical studies. Our
work focuses on the development of a rapid method
of classification, which can be easily used by people
though an appropriate interface and its tools. The
system enables the user to choose, for each tag
which is associated to a content, one of the semantic
meanings drawn from BabelNet. Inserting the tag
“apple”, for instance, the user can select from the
tags menu the most appropriate meaning for his
contents, choosing one of the following senses: «The
apple is the pomaceous fruit of the apple tree, …
etc.», «The Apple Retail Store is a chain of retail
stores owned and operated by Apple Inc., dealing in
computers and consumer electronics», «The apple is
..».
It is clear how this type of classification,
centered on the word meaning in place of the word
form, is particularly useful in the cases in which a
content is classified with polysemantic tags,
connoted with more than one meaning.
In addition to the assignment of semantic tags,
each content is classified by people with one
category and its subcategories.
Content categorization can be done through two
modalities: user can choose one of the Suggested
Categories proposed by the Software, which are
semantically related to the tags inserted, or freely
select a tag and its sub-categories from the dedicated
menu. Categories are extracted from the Open
Directory Project, also known as Dmoz, which
classifies Web sites into 16 main categories, each of
which is divided into several sub-categories.
In addition to the above classification model,
each content can be classified through a meta-data,
also called faceted attribute. The system identifies
four main meta-data, which meet the primary
information needs of user: “ask or give
information”, “test and reviews”, “How to and
guides”, “Exchange, sell or buy”.
The figure below shows the logical components
of the system proposed: the user employs the
bookmarklet to select interesting content (texts,
images, videos) on the World Wide Web; then he
logs in to our Web site in order to classify
semantically the selected content with tags and
categories extracted from the knowledge bases.
Furthermore, the system will allow people to
share the classified content on the main social
networks, such as Facebook and Twitter, supporting
new forms of social crossing interaction and new
ways of relationships among users, based on
common thematic interests.
Figure 1: The logical components of the system proposed.
4 CONCLUSIONS
Our proposal introduces a new method of content
classification, based on an pre-classification
technology that allows an ex-ante disambiguation
which is significantly more reliable than the current
ex-post indexing techniques.
This will allow users to create both a personal
semantic knowledge base, a kind of user's personal
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
316
bookmark, and a collaborative knowledge base
created by the semantic social bookmarking.
The new service is particularly attractive for the
user who feels the need for effective information and
wants to classify his interests, using a simple and
intuitive tool that suggests a disambiguation
taxonomic tree dealing with the typed tags (with the
possibility of reverse path, from general to specific)
(Pani et al., 2012) (Pani, et al., 2013).
Once the content classification is ended, the
service allows user to search and match all the
similarly classified information, which can be
filtered according to optional criteria.
It is a technique which has a high potential for
innovation and market penetration and which will
form the basis for the implementation of an
innovative system which will enable users to:
- select and file any Web content (texts, videos,
images);
- create an archive of content in the personal area;
- meet other community users on the basis of shared
interests and passions;
- find content through the search by title, category
and / or keywords;
- share contents on social networks (Facebook and
Twitter) and interact on them simultaneously
through the innovative social crossing.
5 FUTURE DEVELOPMENTS
The possibility of obtaining an ex-ante
disambiguation of Web content allows users to
communicate and direct advertising with greater
preciseness, opening an interesting frontier of
innovation in this field too.
Thanks to the folksonomic and taxonomic
classification of contents, a more precise contextual
advertising will be achieved. It is well known that
the best advertising addressing induces a more
effective use, and it can become an effective service
when the message is characterized by objective
information elements, and/or interactive inputs,
and/or types of promotional benefits to the user.
Our idea pertains to two main fields:
- to the social and knowledge field, because it
proposes a management tool updated to the real
interests of everyday life, with characteristics of
interoperability between different systems and Web
communities;
- to the economic field because, thanks to its high
disambiguation accuracy, it enables to better address
advertisements and other paid services, which will
be potentially appreciated by people. information
and contacts.
5.1 Further Social Features
In the light of these considerations, our technical
proposal belongs to the social area, which includes
social networks, thematic forums and information
sites whose tools support forms of interaction and
communication among users.
Baseline studies on qualitative and quantitative
dynamics, at global and regional levels, of the major
social networking services reveal a less explosive
phase of growth than in the past, with the
consolidation of the most popular players and
"winners" in the marketplace.
However, at the same time relevant dynamics of
change are occurring in the quantity and types of
user activities, as a result of innovations and new
features incorporated in the main services and in the
market for plugins and third-party applications.
REFERENCES
Moore, R., Lopes, J., 1999. Paper templates. In
TEMPLATE’06, 1st International Conference on
Template Production. SCITEPRESS.
Smith, J., 1998. The book, The publishing company.
London, 2
nd
edition.
Adar, E., Skinner, M., Weld, D. S., 2009. Information
arbitrage across multi-lingual Wikipedia. In:
Proceedings of the Second ACM International
Conference on Web Search and Data Mining,
Barcelona, Spain, pp. 94–103.
Agirre, E., Soroa, A., 2009. Personalizing PageRank for
Word Sense Disambiguation: In: Proceedings of the
12th Conference of the European Chapter of the
Association for Computational Linguistics, Athens,
Greece, pp. 33-41.
Berners-Lee, T., Hendler, J., Lassila, O., 2001. The
Semantic Web. In: Scientific American, pp. 29-37.
Brickley, D., Guha, R. V., 1999. Resource Description
Framework (RDF) Schema Specification. Proposed
Recommendation, World Wide Web Consortium.
http://www.w3.org/TR/PR-rdf-schema
Concas, G., Lisci, M., Pinna, S., Porruvecchio, G., Uras,
S., 2008. Open Source Communities as Social
Networks: an analysis of some peculiar characteristics.
In: 19th Australian Conference on Software
Engineering, ASWEC 2008, pp. 387-391.
Creative Commons, http://creativecommons.org
DBpedia, http://it.dbpedia.org
Delicious, https://delicious.com
DMOZ (Open Directory Project), http://www.dmoz.org
Drupal, https://drupal.org
ClassificationandIndexingofWebContentBasedonaModelofSemanticSocialBookmarking
317
ERC Starting Grant MultiJEDI, Università di Roma “La
Sapienza”. http://lcl.uniroma1.it/multijedi
Facebook, https://www.facebook.com
Fellbaum, C. (Ed.), 1998. WordNet: An Electronic Lexical
Database. MIT Press, Cambridge, MA.
Garcia, A., Szomszor, M., Alani, H., Corcho, O., 2009.
Preliminary Results in Tag Disambiguation using
DBpedia. In: Knowledge Capture (K-Cap'09), First
International Workshop on Collective Knowledge
Capturing and Representation, CKCaR '09, Redondo
Beach, California, USA.
Huynh, D., Mazzocchi, S., Karger, D., 2005. Piggy bank:
experience the semantic web inside your web browser.
In: Proceedings of the 4th international conference on
The Semantic Web, Galway, Ireland.
Joomla!, www.joomla.org
Lehmann, J., Isele, R., Jakob, M., Jentzsch, A.,
Kontokostas, D., Mendes, P. N., Hellmann, S., 2014.
DBpedia - A Large-scale, Multilingual Knowledge
Base Extracted from Wikipedia. In: Semantic Web,
IOS Press, ISSN: 2210-4968.
Lunesu, M. I., Pani, F. E., Concas, G., 2011. An Approach
to manage semantic informations from UGC. In: 3th
International Conference on Knowledge Engineering
and Ontology Development, KEOD 2011, Paris,
France.
Lunesu, M. I., Pani, F. E., Concas, G., 2011. Using a
standards-based approach for a multimedia
knowledge-base. In: 3th International Conference on
Knowledge Management and Information Sharing,
KMIS 2011, Paris, France.
Miller, G. A., 1995. WordNet: A Lexical Database for
English. In: Communications of the ACM, Vol. 38, No.
11, pp. 39-41.
Morsey, M., Lehmann, J., Auer, S., Stadler, C., Hellmann,
S., 2012. DBpedia and the live extraction of structured
data from Wikipedia. In: Program: electronic library
and information systems, Vol. 46 Issue 2, pp.157-181,
ISSN: 0033-0337.
Murgia, A., Concas, G., Marchesi, M., Tonelli, R., 2010.
A machine learning approach for text categorization of
fixing-issue commits on CVS. In: Proceedings of the
2010 ACM-IEEE International Symposium on
Empirical Software Engineering and Measurement.
Navigli, R., Ponzetto, S. P., 2010. BabelNet: building a
very large multilingual semantic network. In
Proceedings of the 48th Annual Meeting of the
Association for Computational Linguistics (ACL), pp.
216-225, Stroudsburg, PA, USA.
Navigli, R., Ponzetto, S. P., 2012. BabelNet: the automatic
construction, evaluation and application of a wide-
coverage multilingual semantic network. In: Journal
Artificial Intelligence, Elsevier Science Publishers Ltd.
Essex, UK, Vol. 193, pp. 217-250.
OmegaWiki, http://www.omegawiki.org
Open Multilingual Wordnet, developed by Francis Bond,
Nanyang Technological University, Singapore,
http://compling.hss.ntu.edu.sg/omw
Pani, F. E., Lunesu M. I., Concas, G., Stara, C., Tilocca,
M. P., 2012. Knowledge Formalization and
Management in KMS. In: 4th International
Conferenze on Knowledge Management and
Information Sharing, KMIS 2012, Barcelona, Spain.
Pani, F. E., Lunesu, M. I., Concas, G., Baralla, G., 2013.
An Approach to Manage the Web Knowlledge. In: 5th
International Conference on Knowledge Engineering
and Ontology Development, KEOD 2013, Algarve,
Portugal.
Passant, A., Laublet, P., 2008. Meaning Of A Tag: a
collaborative approach to bridge the gap between
tagging and Linked Data. In: Proceedings of the
Linked Data on the Web Workshop (LDOW2008) at
the 17th International Semantic Web Conferences
(ISWC 2008). Karlsruhe, Germany, ISSN: 1613-0073.
Spyns, P., de Moor, A., Vandenbussche, J., Meersman, R.,
2006. From folksologies to ontologies: how the twain
meet. In: Proceedings of the 2006 Confederated
international conference on On the Move to
Meaningful Internet Systems: CoopIS, DOA, GADA,
and ODBASE, Vol. 1, pp. 738-755, Springer-Verlag
Berlin, Heidelberg. ISSN: 3-540-48287-3 978-3-540-
48287-1
Suchanek, F. M., Kasneci, G., Weikum, G., 2008. YAGO:
a large ontology from Wikipedia and WordNet. In:
Journal of Web Semantics, Vol. 6, pp. 203–217.
Toral, A., Ferrández, O., Agirre, E., Muñoz, R., 2009. A
study on linking Wikipedia categories to WordNet
synsets using text similarity. In: Proceedings of the
International Conference on Recent Advances in
Natural Language Processing, Borovets, Bulgaria, pp.
449-454 .
Twitter, https://twitter.com
Wikidata, http://www.wikidata.org/
Wikipedia, http://www.wikipedia.org
Wiktionary, http://www.wiktionary.org/
WordNet, http://wordnet.princeton.edu
Wordpress, https://wordpress.com
Wu, F., Weld, D., 2007. Automatically semantifying
Wikipedia. In: Proceedings of the Sixteenth ACM
Conference on Information and Knowledge
Managemen t, Lisbon, Portugal, pp. 41-50.
Zhong, Z., Ng, H. T., Chan, Y.S., 2008. Word Sense
Disambiguation using OntoNotes: An empirical study.
In: Proceedings of the 2008 Conference on Empirical
Methods in Natural Language Processing, Waikiki,
Honolulu, Hawaii, pp. 1002-1010.
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
318
