ABCT: The Activity based Contextual Tagging Ontology
Grégory Bourguin and Arnaud Lewandowski
University of Lille Nord de France, ULCO, LISIC, France
Keywords: Tagging, Context, Activity, Viewpoint, Ontology, MUTO, PROV.
Abstract: A large amount of applications now includes tagging mechanisms that have proven efficiency to organize,
navigate through, retrieve, and discover online resources. However, despite the valuable research work done
to improve these solutions, the literature shows that a further step has to be done in order to better consider
the contexts in which tagging actions occur. In this paper, we define a list of elements constituting a tagging
context that should be considered in order to better give access to the knowledge shared through users’
taggings. We propose an ontological model named ABCT (Activity Based Contextual Tagging) for describing
these contexts. ABCT takes benefits from the many research in tagging ontologies and that are synthetized in
MUTO (Modular Unified Tagging Ontology). ABCT marries MUTO and PROV (Provenance) concepts to
facilitate the description of tags and tagging contexts, essentially through to the notions of Tagging and
Activity.
1 INTRODUCTION
In the past decade, tagging systems have become an
essential part of a wide range of applications
(knowledge-management, social media, repositories,
online stores, etc.). One of the reasons of this
infatuation is that, in our world where the number and
variety of resources (information, materials,
software…) are constantly and rapidly growing,
researchers have shown that tagging systems can
really help in organizing, navigating through, and
retrieving them (Ames and Naaman, 2007; Oleksik et
al., 2009). Moreover, tagging systems have been
successfully used to let end-users (by opposition to
resource developers and/or domain experts)
themselves organize this plethora. Researchers have
shown that this mechanism is really interesting since
tags reflect their creator’s experience, and tagging a
resource is sharing knowledge about it (Saab, 2010).
In tagging systems, end-users’ knowledge shared
through tags is expected to help other users in better
finding, understanding and selecting resources
according to their own specific needs (Singer et al.,
2013).
Many research work has been realized to enhance
tagging systems. In a previous paper (Bourguin and
Lewandowski, 2017), we proposed a litterature
review that explores studies and solutions dedicated
to folksonomies (Knerr, 2006; Cernea et al., 2008;
Saab, 2010) and ontologies (Kotis and Vouros, 2006;
Dong et al., 2015; Garcia-Silva et al., 2014;
Zhitomirsky et al., 2017), while trying to better
understand the essence of tags. A tag carries
semantics that provides meaning about a resource.
However, information is only useful to the extent that
other users make sense of the content in the same way
(Golder and Huberman, 2006). A tag reflects it’s
creator’s knowledge, but knowledge can be fully
understood only while considering the context it
comes from (Ning and O'Sullivan, 2012). This
explains why most tagging solutions have proposed
to enhance tags by linking them to information related
to their creation context.
Following this trend, we also proposed the basis
of a new tagging framework (Bourguin and
Lewandowski, 2017). However, our analysis leaded
us to consider the notion of context in a wider way
than in the previous solutions by linking tags not only
to their creator (or creator’s intention), but to their
whole creator’s activity, thus potentially providing
much more contextual information.
In this paper, we consolidate our approach by
defining the Activity Based Contextual Tagging
(ABCT) ontology. ABCT is designed to take benefits
from most of the previous propositions in Tagging
ontologies and that were synthesized in the Modular
Unified Tagging Ontology (MUTO, Lohmann et al.,
2011). ABCT extends MUTO by merging its
200
Bourguin, G. and Lewandowski, A.
ABCT: The Activity based Contextual Tagging Ontology.
DOI: 10.5220/0006953902000207
In Proceedings of the 10th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2018) - Volume 3: KMIS, pages 200-207
ISBN: 978-989-758-330-8
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

concepts with the W3C PROV ontology (W3C, 2013)
which’s purpose is to describe entities and
provenance information synthesized in the concept of
Activity. To our knowledge, ABCT is the first tagging
ontology explicitely anchoring tags in their creator’s
activity, i.e. where they were created and used, and
thus enabling to capture more information about the
context in which they really make sense.
In the first part of the paper, we define elements
that should be considered while capturing tagging
contexts, because they bring information that allows
to assess them and to understand a user’s particular
viewpoints. The second part of the paper presents
ABCT, the ontological model we propose to describe
tagging in context, which can be viewed as a
dedicated marrying of the MUTO and PROV data
models. Finally, we illustrate ABCT’s main benefits
with an example showing how this framework can
help to capture contextual users’ taggings in a global
multi-viewpoint ontology.
2 TAGGING CONTEXT
While studying strengths and weaknesses of
folksonomies, many researches have shown that
tagging context is crucial for better understanding
tags (Ning and O'Sullivan, 2012). Qassimi et al.
(2016) underline that tags synonymy (multiple tags
holding the same meaning) is not transitive but
fundamentally context dependent. They also recall
that polysemy in one of the central issues in the
psychology of word meaning, and that a polysemous
word cannot be fully understood if considered out of
context.
In fact, many tagging solutions propose to link
tags to some entities related to their creation context.
In most cases, this information is directly related to
the creator of the tag. For example, Newman’s TAGS
(Newman, 2005) proposed to link tags to their tagger
through the tagging concept, and NiceTag (Monnin et
al., 2010) proposed to represent the intention of the
tag’s creator.
Most of these initial propositions were
synthesized in MUTO (Lohmann et al., 2011), a
tagging ontology we will further describe in the next
part of this paper. However, we argue that this
information, even if necessary, is not sufficient to
fully understand tags and tagged resources. Our
assumption is that we need to enlarge the notion of
context. We propose here the list of elements that we
consider as the context that can help in understanding
a tagging action.
2.1 Contextual Elements
The following elements are those usually already
represented in existing tagging systems:
• The resource. Obviously, knowing which
resource is concerned by a tag helps in better
understanding it. For example, the label ‘pink’
can have several meanings. Knowing that this
label has been associated with a photo of the
famous singer, the ambiguity is weakened.
• The other tags on the same resource. The tag
‘beach’ usually triggers some deep blue sea and
palm trees images in the mind. If we see the other
tags ‘Battle of Dunkirk’ and ‘Second World
War’ on the same entity, it gives us more
knowledge about this ‘beach’ which does not
refer to some paradisiac island, but to the historic
place where tragic events occurred.
• The tag’s creator. The ‘Java’ tag may refer to a
town, a France originated dance, or even a
chicken. Knowing that a well-known software
programmer created this tag lets understand that
the targeted resource has more probably links
with the famous programming language.
The following elements are related to the activity
in which tagging occurred. Even if some tagging
systems provide some clues about the taggers’
activity, like the localization of the tagging action
(Qassimi et al., 2016), they do not consider activity
and its related elements in the large as we do here:
• The activity in which the tagging occurs.
Finding the tag ‘Head’ on a Christmas ball may
be somehow confusing. As we will show in our
example described in part 4, discovering that this
tag has been created in an activity dedicated to
the creation of a decorative unicorn for a child
bedroom levers the ambiguity.
• The other resources involved in the activity. A
resource is rarely used alone. Knowing which
other resources are associated with a particular
one can help to assess this latter and the tag(s) put
on it. For example, a tag ‘model’ describing an
entity in an activity that involves a camera and
spotlights will not trigger the same meaning than
another entity associated with a programming
language and a database server.
• The tags on the other resources. Let’s consider
once again our Christmas ball tagged with
‘Head’. Even without considering the activity
name, this tag is more easily understood if
discovered with other specific resources tagged
‘horn’, ‘body’, ‘legs’ and ‘rainbow’.
ABCT: The Activity based Contextual Tagging Ontology
201

• The tags on the activity can be used to inform
about the activity type and provide useful
contextual information. In our previous work
(Bourguin and Lewandowski, 2015, 2017), we
considered MMORPG players (Massive
Multiplayer Online Role Playing Game) sharing
their character’s build. Each build (an
assemblage of game resources) is relevant for a
specific activity type, and a each type is usually
described by tags like ‘tanking’, ‘healing’, ‘dps’
(damage dealing). Players sometimes tag
resources as ‘OP’ (Over Powered). Such
information is really impacting for other players,
but ‘OP’ does not give much information if not
considered in the context of an activity type:
indeed, a resource can be ‘OP’ for ‘tanking’, but
highly inadvisable for a ‘dps’ activity.
• The tags on the creator can inform about his/her
profile, his/her role(s) in the activity, and this
knowledge could give more sense to the
considered tag. For example, if a user is tagged
‘Web designer’, we understand that the tag
‘Head’ s/he put on a particular resource stands
for its position on the web site, and not for the
upper part of the human body.
• The other actors in the activity. Let’s consider
the tag ‘to read’ put on a particular resource. If
only one user is involved in a private activity, we
may deduce that s/he created this tag for
her/himself, as a reminder. At the opposite, if ‘to
read’ is created by a teacher in an activity
involving several students, we understand that
this tag stands for a reading recommendation.
• The other activities a tagger is involved in can
help in assessing a tag. For example, knowing the
many projects in which a programmer is
involved in, and her/his contributions, offers an
overview of her/his skills and experience, which
may be concentrated around the Java language:
this can lead to better understand why s/he tagged
as ‘Best’ a GUI Java Framework, while C++
specialists would certainly have tagged ‘Best”
another one. Indeed, from its surrounding
activities, we implicitly understand here ‘Best’ as
‘Best for Java’ instead of ‘Best of all’.
2.2 Capturing and Sharing Viewpoints
As described in the previous part, we think that a
tagging system would gain benefits from linking tags
to numerous contextual elements. Each entity is
linked to many other entities (e.g. all the tags
associated by all the system’s users with a particular
resource in a folksonomic fashion), but some links are
stronger since they represent a particular context that
can help to better understand the act of tagging, and
then the associated meaning. Focusing on these
particular stronger relations between entities in a
particular context offers what we call a viewpoint.
The viewpoint notion is more and more identified
as essential by researchers interested in sense making.
Indeed, providing meaning about things is always
sharing a viewpoint. As recalled by Bénél and
Lejeune (2009), meaning of things is always plural
and trying to provide a unique definition is thus
problematic in essence. ZhitomirskyGeffet et al.
(2017) underline that even ontologies defined by
domain experts and expected to provide a unique and
agreed shared definition about domains entities, can
only be considered as experts’ viewpoints. They also
underline that different experts rarely share the same
viewpoint, and even more that an expert’s viewpoint
rarely matches the ontology’s users one. This
certainly explains why a large part of today’s research
in ontologies is dedicated to ontologies alignment, a
discipline trying to create bridges between different
ontologies representing the same domain, but defined
by different experts, thus providing different
viewpoints. This also certainly explains why a new
trend has emerged in this research area while
introducing the need for multi-viewpoints ontologies
(Kotis and Vouros, 2006) (Pinto et al., 2009)
(Zhitomirsky et al., 2017). Indeed, as reported by
Zhou and Bénel (2008), a system for helping sense
making should let users distinguish and compare
viewpoints: once interpretation conflicts permit to
distinguish different viewpoints, people are then able
to choose and/or create their own.
As a result, we think that a tagging system has to
provide means to retrieve the viewpoint a set of
entities are participating in, to focus on a particular
one for better understanding, and to browse and
compare them. This approach is really close to the
one developed in the frame of Hypertopic (Cahier and
Zacklad, 2006) and its associated technologies
(Cahier et al., 2013). The Hypertopic framework lets
users define a Corpus in which Items can be
associated with a Topic (tag) in the context of a
Viewpoint (a set of Topics characterizing Items). The
main difference between Hypertopic and our
approach relies in our framework definition since we
choose to use the Activity (and constituting elements)
as a central concept to provide context for Tagging.
The main motivation is to more directly consider that
tagging is an action that is performed by an actor in
the context of an activity, i.e. a motivated aggregate
of actors, resources, etc. In our approach, Activity and
Tagging are used together to capture and share
KMIS 2018 - 10th International Conference on Knowledge Management and Information Sharing
202

viewpoint(s). The resulting framework has been
defined as an ontology we called ABCT.
3 ABCT ONTOLOGY
ABCT (Activity Based Contextual Tagging) is an
evolution of our previous work (Bourguin and
Lewandowski, 2015, 2017). In this previous
approach, the main idea was to let users describe their
viewpoints into Personal Ontologies (PO). Indeed,
each PO was itself a separate ontology in which tags
were not instances of a tagging concept, but were
themselves concepts (i.e. instances of owl:Class).
Users were able to describe their own understanding
of the world in their own ontologies, while directly
taking benefits from ontological mechanisms like
consistency checking and inference. Our application
framework was designed to let end-users unfamiliar
with semantic technologies create their own owl
ontologies. Experiments have shown interesting
results, letting users from different domains
(scrapbooking, MMORPG, e-learning) share their
experience. However, this approach showed
limitations too mainly due to the fact that, each
viewpoint being a PO, the application generated
numerous separate ontologies.
In ABCT, we aim at providing a single global
ontology containing all the user’s different
viewpoints, thus facilitating the global querying that
will serve our many purposes like exploration,
viewpoints comparison, recommendation and so on.
The two main concepts driving this research are
Tagging and Activity. Instead of creating a new model
from scratch, we decided to take benefits from
recognized research results in modelling these two
concepts. Representing Tagging is the purpose of
MUTO, and Activity description is the motivation of
W3C’s PROV. ABCT stands as the merging of
MUTO and PROV in a new ontology designed to
support our needs.
3.1 MUTO
Trying to cope with the limitations of the first
folksonomic systems, several tagging ontologies have
been developed. Each of these ontologies proposed a
variation of the semantic representation of
folksonomies and, more precisely, a model for
representing a tagging while enhancing the tag
concept. As reported by Lohmann et al. (2011), the
large number of these tagging ontologies made it
difficult for developers to find the best ontology that
meets their need. The authors thus proposed the
Modular Unified Tagging Ontology (MUTO), a
unification of the existing tagging models. MUTO
provides a tagging ontology designed to combine the
best of the nine most recognized tagging ontologies,
and this is why we decided to use it as a foundation in
ABCT for representing the Tagging concept.
Briefly described, in MUTO, users (single user or
group) instances of sioc:UserAccount, create some
muto:Tagging. Each muto:Tagging is linked to a
rdfs:Resource and can contain ordered instances of
muto:Tag. Each tag is unique (even if multiple tags
have the same string as rdfs:label) and can be
associated with a meaning (a rdfs:Resource like a
concept in another ontology). More information can
be described like tag hierarchy (a muto:Tag being
subclass of skos:Concept), private tagging, auto
tagging, creation and modification dates, etc. The full
descriptions of MUTO can be found in (Lohmann et
al., 2011).
As we can see, MUTO provides an interesting
basis for managing users’ tagging (and tags). Some
contextual information can be described: mainly the
user’s account that created the tagging. However,
MUTO does not provide entities facilitating the
description of the tagging context as we introduced it
previously. In our approach, the context for tagging
actions is the Activity, a concept that doesn’t exist in
MUTO, but that is central in W3C’s PROV.
3.2 PROV
The PROV data model is a standard proposed by the
World Wide Web Consortium (W3C) to represent the
provenance and history of data on the web. The W3C
defines provenance as “a record that describes the
people, institutions, entities, and activities involved in
producing, influencing, or delivering a piece of data
or a thing.” (W3C, 2013). According to the W3C,
capturing and representing the provenance of
information can help users to understand it, to decide
whether to trust it, or to know how to integrate it
Figure 1: PROV data model core concepts (W3C, 2013).
ABCT: The Activity based Contextual Tagging Ontology
203

Figure 2: ABCT ontological data model, based on MUTO and PROV.
elsewhere. These considerations clearly correspond
to our needs concerning the elements that constitute
the context of a tag and that could help in assessing it.
Furthermore, the PROV data model has been found
usable and useful for end-users (Bachour et al., 2015).
Figure 1 shows the core concepts of the PROV
data model, which are described with more details in
(W3C, 2013) and (Moreau et al., 2015): prov:agents
participate in prov:activities that can use or produce
prov:entities. These entities describe digital, physical
or other things (documents, objects, web sites, etc).
Thanks to these three core concepts and their
relationships, PROV can model the creation
(prov:wasGeneratedBy) and the usage (prov:used) of
resources, the derivation of resources from other
resources and the versioning of resources
(prov:wasDerivedFrom), humans or other things
involved in activities (prov:wasAssociatedWith), or
being responsible for some entity
(prov:wasAttributedTo). Furthermore, PROV data
model offers the means to refer to several other
concepts such as time, location, role, and plan. All
these concepts can help us to describe the context in
which a tagging action occurred.
3.3 ABCT
ABCT is designed to ease the instantiation of
(MUTO) Tagging(s) framed in the context of (PROV)
Activity(ies). For this purpose, ABCT classes and
properties (see Figure 2) inherit from MUTO and/or
PROV classes and properties. For example, the
abct:Tagging class is defined both as a subclass of
muto:Tagging and abct:Resource, which itself is a
subclass of prov:Entity. Doing so, a tagging can be
described as an abct:Resource being part of
(abct:wasUsed) and/or generated by
(abct:wasGeneratedBy) some abct:Activity (a
subclass of prov:Activity). We can also specify that
this tagging has been created (abct:hasCreator : sub-
property of both prov:wasAttributedTo and
muto:hasCreator) by an abct:Agent (subclass of
prov:Agent, and sioc:UserAccount – cf. MUTO) that
itself can be an abct:agentOf some abct:Activity.
Inspired by MUTO, an abct:Tagging may contain
multiple instances of abct:Tag; this explains why
abct:Tagging is also defined as a subclass of
prov:Collection, and why the abct:hasTag object
property inherits from muto:hasTag and
prov:hadMember. One can notice that an
abct:Tagging is targeting (abct:hasResource) an
rdfs:Resource, which is the superclass of all of the
previously described entities. As a result, this model
enables to (contextually) tag any resource, but also
agents and activities. In fact, even taggings and tags
could themselves be tagged by agents in the same or
other activities, thus providing their viewpoint
concerning a tagging or tag: an information that can
for example help in creating collaborative features
supporting the building of shared viewpoint(s) like in
HCOME (Kotis and Vouros, 2006) and Collaborative
Protégé (Tudorache et al., 2008).
Instances of this model allow to describe tagging
performed by some (types of) agents in some (types
<<owl:Class>>
prov:Activity
<<owl:Class>>
muto:Tag
<<owl:Class>>
muto:Tagging
<<owl:Class>>
rdfs:Resource
<<owl:Class>>
prov:Agent
<<owl:Class>>
sioc:UserAccount
<<owl:Class>>
prov:Entity
<<owl:Class>>
prov:Collection
<<owl:ObjectProperty>>
abct:hasAgent
inverse(agentOf)
<<owl:ObjectProperty>>
abct:wasGeneratedBy
inverse(generated)
<<owl:ObjectProperty>>
prov:wasGeneratedBy
inverse(prov:generated)
<<owl:ObjectProperty>>
prov:hadMember
inverse(prov:wasMemberOf)
<<owl:ObjectProperty>>
muto:hasMeaning
inverse(muto:meaningOf)
<<owl:ObjectProperty>>
abct:hasResource
inverse(resourceOf)
<<owl:ObjectProperty>>
muto:hasResource
inverse(muto:resourceOf)
<<owl:ObjectProperty>>
prov:wasAttributedTo
inverse(prov:contributed)
<<owl:ObjectProperty>>
muto:hasCreator
inverse(muto:creatorOf)
<<owl:ObjectProperty>>
abct:wasUsedBy
inverse(used)
<<owl:ObjectProperty>>
prov:wasUsedBy
inverse(prov:used)
<<owl:ObjectProperty>>
muto:hasTag
inverse(muto:tagOf)
<<owl:Class>>
abct:Resource
<<owl:Class>>
abct:Activity
<<owl:Class>>
abct:Agent
<<owl:ObjectProperty>>
abct:hasMeaning
inverse(meaningOf)
<<owl:ObjectProperty>>
abct:hasTag
inverse(tagOf)
<<owl:Class>>
abct:Tag
<<owl:Class>>
abct:Tagging
<<owl:ObjectProperty>>
prov:wasAssociatedWith
inverse(prov:wasAssociateFor)
<<owl:ObjectProperty>>
abct:hasCreator
inverse(creatorOf)
<<owl:ObjectProperty>>
abct:wasAttributedTo
inverse(abct:contributed)
KMIS 2018 - 10th International Conference on Knowledge Management and Information Sharing
204

Figure 3: Partial example describing several tagging contexts using ABCT ontological data model.
(* Emily appears twice in this figure for readability, but the 2 boxes represent the same entity).
of) activities, and its implementation (we realized
with Apache Jena) makes it possible to retrieve
specific taggings and their associated contextual
information through SPARQL queries.
As it was underlined in the previous paragraphs,
MUTO and PROV ontologies both provide means for
describing more detailed information like tagging
creation and modification date, tags ordering and
hierarchy (inherited from SKOS through MUTO),
and the possibility to keep track of the evolution of
the many entities through PROV (e.g. specifying that
a tag or tagging was derived from another one). In
other words, creating and specifying ABCT entities
and properties allows to describe contextual tagging,
but it also enables to provide more detailed
information concerning each facet: (MUTO) tagging
and (PROV) activity.
4 SAMPLE APLICATION
Figure 3 briefly illustrates how the ABCT framework
can be used to model three specific viewpoints related
to a same resource. It is inspired by our nascent work
in applying ABCT in the development of new
functionalities for a French company. This company
mainly sells decoration supplies (furniture, materials,
etc.) in around twenty French stores, and also
provides an online shop in which its (decoration)
designers and customers can share pictures and
explanations about what they produced while using
some of the shop’s goods in a DIY (Do It Yourself)
trend. For reasons of space and conciseness, this
example uses information extracted from real articles
posted on the website and sample ideas mixed
together for quickly showing the problems related to
tags’ semantics, and the solution proposed by ABCT
in providing the needed contextual information.
The example focuses on a simple shop’s good
presented with a picture: a plastic transparent
Christmas Ball. One can notice that, in this small
example and following a classical folksonomic
approach, this resource would be presented with all
the tags: Head, Blue, Transparent and Container.
Such description of a Christmas ball can be somehow
intriguing and confusing: how can it be a head, a
container (for a brain?), at the same time blue and
transparent? This is where the details captured
through the ABCT framework can help.
In this example, ABCT enables to know that an
abct:Agent named Emily is involved in an
abct:Activity named Unicorn making, a DIY activity
aiming at building a unicorn for decorating a child
bedroom and by assembling several materials. In this
activity, Emily uses the Christmas ball as the
unicorn’s Head, puts some Blue paint on it, and other
resources that do not appear in the figure 3 for
readability. She did not use the Container tag in this
activity: it appears meaningless in this context. She
also put the tag Skin Color on the blue paint: blue is
abct:used
<<abct:Agent>>
Emily*
<<abct:Resource>>
Christmas Ball
<<abct:Activity>>
Unicorn making
<<abct:Tagging>>
<<abct:Tag>>
Head
<<abct:Agent>>
Bill
<<abct:Activity>>
December product
promotion
<<abct:Tagging>>
<<abct:Tag>>
Head
<<abct:Agent>>
Emily*
<<abct:Activity>>
My Christmas
Tree
<<abct:Tagging>>
<<abct:Tag>>
Transparent
abct:creatorOf
abct:agentOf
abct:agentOf
abct:used
abct:used
abct:used
abct:used
abct:hasTag
<<abct:Resource>>
Polystyrene
Beads
abct:hasResource
abct:hasResource
abct:used
abct:creatorOf
abct:agentOf
abct:used
abct:hasResource
<<abct:Tag>>
Blue
abct:hasTag
<<abct:Tagging>>
<<abct:Tag>>
Snowflakes
abct:creatorOf
<<abct:Resource>>
Blue Paint
<<abct:Tagging>>
<<abct:Tag>>
Skin Color
abct:hasTag
abct:hasResource
abct:creatorOf
abct:used
abct:used
<<abct:Resource>>
Vendor
Website
abct:used
<<abct:Tag>>
Container
abct:hasTag
abct:used
abct:hasTag
abct:hasResource
abct:wasAttributedTo
abct:creatorOf
The front part of the body in animals ;
Contains the face and brains
(WordNet)
abct:hasMeaning
The top of something
(WordNet)
abct:hasMeaning
ABCT: The Activity based Contextual Tagging Ontology
205

not an awaited colour for skin, but it is actually
meaningful in the Unicorn making activity.
Emily performed another activity named My
Christmas tree where she also used the same
Christmas Ball, but this time as a Transparent
Container filled with some Polystyrene Beads tagged
as Snowflakes. As we can see, our same abct:Agent
adopted a somewhat different viewpoint on the
Christmas Ball while associating it with different
resources and for another purpose.
Finally, the example shows that this Christmas
ball is also characterized from a web designer
viewpoint, in an activity named December product
promotion. Bill is in charge of organizing the vendor
website and he put the tag Head on the Christmas
Ball, to denote the fact that this article has to be used
in the heading of the website. One can notice that both
Emily and Bill used the Head tag on the Christmas
Ball. However, using the abct:hasMeaning property,
this same label can be associated with different
meanings depending on the adopted viewpoint.
This example shortly illustrates some of the many
problems that may rise when the tagging contexts are
not explicit. It also exemplifies how the contextual
elements identified in part 2.1 can be made explicit
through the ABCT framework. Querying the
populated ontology can help in discovering, browsing
and selecting viewpoints, and then assessing and
better understanding the many tags as they were
thought from their creator’s viewpoint.
5 CONCLUSIONS
Despite of the large number of solutions using a
tagging system to let users share their knowledge
about diverse resources, the literature shows that
tagging models still miss important information.
Most researchers report that tags cannot be fully
understood if disconnected from the context in which
the corresponding tagging actions occurred. Even if
evolving research in tagging ontologies has proposed
new models to capture more contextual information,
we think that considering a larger notion of context
can enhance these models.
In our approach, the context for tagging action is
the tagger’s activity that frames the many entities
participating to a task performance. Inspired by
previous research, we listed the main elements that
provide contextual information for a better
understanding of tags and tagged entities. Our main
idea is that each particular activity offers a viewpoint,
and exploring a viewpoint gives access to the specific
set of elements that provides the context needed to
understand each other.
Building on these results, we proposed a new
framework founded on the Tagging and Activity main
concepts. Our proposition aims at providing a
contextual tagging ontological model that facilitates
the building of ontology that captures and allows
exploring viewpoints. For this purpose, we proposed
the ABCT ontology, a marriage of MUTO – the
synthesis of most recognized Tagging ontologies –
and PROV – W3C’s ontology for representing
provenance thanks to the notion of Activity.
With a small sample application, we underlined
the main features and possibilities provided by ABCT
for capturing contextual tagging. Due to a lack of
space, we did not explore here dimensions concerning
specific viewpoints connections and that can be
represented by specific links between activities, like
the recursive structure that lets define global or sub
activities (allowing to represent different viewpoints
participating in a global and maybe cooperative
activity).
ABCT’s implementation is only at its beginning.
We already created an OWL representation of ABCT
with Protégé, and put it in action in a JEE REST
server. Our first applications use Apache Jena for
ontology building (capturing viewpoints) and
SPARQL for querying (exploring and exploiting
viewpoints). These early experiments in supporting
knowledge builders and users involved in the
different application domains we briefly mentioned in
this paper (e-learning, software development, e-store,
and gaming) are actually very promising.
REFERENCES
Ames, M., & Naaman, M. (2007). Why we tag: motivations
for annotation in mobile and online media. In
Proceedings of the SIGCHI conference on Human
factors in computing systems, pp. 971-980.
Bachour, K., Wetzel, R., Flintham, M., Huynh, T. D.,
Rodden, T., & Moreau, L. (2015). Provenance for the
people: an HCI perspective on the W3C PROV
standard through an online game. In Proceedings of the
33rd Annual ACM Conference on Human Factors in
Computing Systems, pp. 2437-2446.
Bénel, A., & Lejeune, C. (2009). Humanities 2.0:
documents, interpretation and intersubjectivity in the
digital age. In International journal of web based
communities, 5(4), pp. 562-576.
Bourguin, G., & Lewandowski, A. (2015). Using online
personal ontologies to share experience about web
resources. In Proceedings of Information Systems, pp.
177-184.
KMIS 2018 - 10th International Conference on Knowledge Management and Information Sharing
206

Bourguin, G., & Lewandowski, A. (2017). A Framework
for Experience Sharing Through Contextual Tagging.
In Proceedings of the 9th International Joint
Conference on Knowledge Discovery, Knowledge
Engineering and Knowledge Management – Vol. 3:
KMIS, pp. 205-211.
Cahier, J.-P., & Zacklad, M. (2006). The Agoræ/
Hypertopic approach. In Proceedings of the workshop
on Indexing and Knowledge in Human Sciences,
Nantes, June 26-28, 2006, pp. 66-70.
Cahier, J. P., Bénel, A., & Salembier, P. (2013). Towards a
“non-disposable” software infrastructure for
participation. In Interaction Design and Architecture(s)
Journal, 18, pp. 68-83.
Cernea D.A., Del Moral E., Gayo J.E.L. (2008). SOAF:
Semantic indexing system based on collaborative
tagging. In Interdisciplinary Journal of E-Learning and
Learning Objects 4, pp. 137-149.
Dong H., Wang W., Liang H.N. (2015). Learning
Structured Knowledge from Social Tagging Data: A
Critical Review of Methods and Techniques. In Proc.
of 8th SocialCom, Chengdu, China.
Golder S., & Huberman B.A. (2006). Usage patterns of
collaborative tagging systems. In Journal of
Information Science, 32(2), pp. 198–208.
Garcia-Silva A., Garcia-Castro L. J., A.García, Corcho O.
(2014). Social tags and Linked Data for ontology
development: A Case Study in the Financial Domain.
In Proc. of WIMS14, Thessaloniki, Greece.
Knerr, T. (2006). Tagging ontology-towards a common
ontology for folksonomies.
Kotis K., Vouros G. A. (2006). Human-centered ontology
engineering: The HCOME methodology. In Knowledge
and Information Systems, 10(1), pp. 109-131.
Lohmann, S., Díaz, P., & Aedo, I. (2011). MUTO: the
modular unified tagging ontology. In Proceedings of
the 7th Int. Conf. on Semantic Systems, pp. 95-104.
Monnin A., Limpens F., Gandon F., Laniado D. (2010).
Speech acts meet tagging: NiceTag ontology. In Proc.
of I-SEMANTICS’10, pp. 31:1–31:10. ACM.
Moreau, L., Groth, P., Cheney, J., Lebo, T., & Miles, S.
(2015). The rationale of PROV. In Web Semantics:
Science, Services and Agents on the World Wide Web,
35, pp. 235-257.
Newman, R. (2005). Tag ontology design.
http://www.holygoat.co.uk/projects/tags/
Ning, K., & O'Sullivan, D. (2012). Context modeling and
measuring for context aware Knowledge Management.
In International Journal of Machine Learning and
Computing, 2(3), 287.
Oleksik, G., Wilson, M. L., Tashman, C., Mendes
Rodrigues, E., Kazai, G., Smyth, G., ... & Jones, R.
(2009). Lightweight tagging expands information and
activity management practices. In Proceedings of the
SIGCHI Conference on Human Factors in Computing
Systems, pp. 279-288.
Pinto H.S, Tempich C., Staab S. (2009). Ontology
engineering and evolution in a distributed world using
DILIGENT. In Handbook on Ontologies, pp 153-176.
Qassimi, S., Hafidi, M., & Lamrani, R. (2016). Enrichment
of ontology by exploiting collaborative tagging
systems: a contextual semantic approach. In
International Conference on Systems of Collaboration
(SysCo), pp. 1-6.
Saab D. J. (2010). The Ontology of Tags. iConference
2010.
Singer, L., Figueira Filho, F., Cleary, B., Treude, C., Storey,
M. A., & Schneider, K. (2013). Mutual assessment in
the social programmer ecosystem: An empirical
investigation of developer profile aggregators. In
Proceedings of the 2013 conference on Computer
supported cooperative work, pp. 103-116.
Tudorache T., Noy N. F., Tu S., Musen M. (2008).
Supporting Collaborative Ontology Development in
Protégé. In Proceedings of the 7th International
Conference on The Semantic Web (ISWC'08), Springer-
Verlag, Berlin, Heidelberg, pp. 17-32.
W3C (2013). Prov-DM: The PROV Data Model. W3C
Recommendation, World Wide Web Consortium.
https://www.w3.org/TR/prov-dm/
Zhitomirsky-Geffet, M., Erez, E. S., & Judit, B. I. (2017).
Toward multiviewpoint ontology construction by
collaboration of nonexperts and crowdsourcing: The
case of the effect of diet on health. In Journal of the
Association for Information Science and Technology,
68(3), pp. 681-694.
Zhou, C., & Bénel, A. (2008). From the crowd to
communities: New interfaces for social tagging. In
From CSCW to Web 2.0: European Developments in
Collaborative Design Selected Papers from COOP08.
ABCT: The Activity based Contextual Tagging Ontology
207