WISHFUL AND WISDOM AWARE COMPOSING
A Full User-centric Approach to Create NGM
Luis Javier Suarez Meza
1
, Luis Antonio Rojas Potosi
1
, Juan Carlos Corrales
1
and
Oscar Rodriguez Rocha
2
1
Grupo de Ingenieria Telematica (GIT), Universidad del Cauca, Popayan, Colombia
2
Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Torino, Italy
Keywords:
Wishful and Wisdom Aware Composing, User-centric, Next Generation Mashups.
Abstract:
Nowadays, it’s possible to find huge amounts of steadily increasing web resources. Service composition lan-
guages and tools are widely used for creating compositions of multiple services in order to meet the user’s
requirements in the cases when a single Web service does not perform a required task. However, despite the
relative intuitiveness of currently the available tools, they still require a careful manual assembly of a com-
position flow by the end-user, thus requiring some technical knowledge on the functioning of each individual
component. We provide a full user-centric approach to create NGMs (Next Generation Mashups) interactively
through W2AC (Wishful and Wisdom-Aware Composing): First, end-user wishes are considered, then a com-
position knowledge is extracted from existing reputable mashups created by skilled users. Thus, end-users are
able to easily generate their own NGM in a fully customized fashion.
1 INTRODUCTION
The Wishful and Wisdom concepts are the founda-
tions of our work. Generally, in many related works
(Chowdhury et al., 2010; Riabov et al., 2008; De An-
geli et al., 2011; Casati, 2011), these concepts are
considered in a separated fashion. We take advantage
of the potential of this combination (users’ desire +
collective knowledge), to embrace a resource compo-
sition focused on the end-user. To explain our W2AC
vision, some concepts are introduced.
Our approach is based on BDI systems, which has
a modal component to reason about propositional at-
titudes: beliefs, desires and intentions (BDI) (Das-
tani and Steunebrink, 2009). In this sense, we cover
these aspects under the Wishful concept. Thus, we
aim to determine what the user really wants from an
explicit request made in natural language (Pedraza
et al., 2011), since it is difficult to proposetothe user a
service that meets his/her search requirements, with-
out knowing the true meaning of what he/she really
wants.
Wisdom is a concept closely related to the
paradigm of collective intelligence (CI) (Agarwal
et al., 2010; Dalal, 2008). Generally, some ap-
proaches that have considered CI, argue that the users
do not need a high level of expertise in service cre-
ation platform for composing mashups, if they have
an adequate assistance, advice, or help by another
users, who have previously solved the same (or simi-
lar) problem (Szuba et al., 2011; Maries and Scarlat,
2011).
On the other hand, many research efforts have
been done on automatic composition, especially
within the AI planning and Semantic Web commu-
nities. Other work uses process models or formal
representation (e.g., Graphs) (Maaradji et al., 2011;
Chowdhury et al., 2011). Further, current approaches
about composition workon resources of the same type
(i.e., just on Web services, like BPEL) (Riabov et al.,
2008). The review on the state of the art shows there
are currently no approaches that dare to make a com-
bination of the great diversity of existing resources.
Wasting the diversity of available resources and limit-
ing the emergence of novel and interesting resources.
In this context, we propose a combination of web re-
sources called NGM. NGM is an hybrid integration
of different types of available resources created by
end-users. We provide an approach to create NGM’s
interactively through W2AC: First, end-user wishes
are considered, then a composition knowledge is ex-
tracted from existing reputable mashups or different
resources created by other more technically skilled
users.
803
Suarez Meza L., Rojas Potosí L., Corrales J. and Rocha O..
WISHFUL AND WISDOM AWARE COMPOSING - A Full User-centric Approach to Create NGM.
DOI: 10.5220/0003940108030806
In Proceedings of the 8th International Conference on Web Information Systems and Technologies (WEBIST-2012), pages 803-806
ISBN: 978-989-8565-08-2
Copyright
c
 2012 SCITEPRESS (Science and Technology Publications, Lda.)

2 THE USERS’ DESIRES
COMPOSER: W2AC
This section presents the formal definition of the main
components of an NGM: a query model that describes
the formal user request and a Mashup model to de-
rive the semantic description of an NGM, based on
the descriptions of its individual components. A fun-
damental characteristic of our model is that it captures
not only the semantics of inputs/outputs (and its func-
tional dependency), and operations, but also the se-
mantics of control operator structures (i.e., composi-
tion structure patterns).
2.1 Query Model
Currently there is no a model that represents the for-
mal request of end-users, therefore, below we present
our proposal for this concern.
2.1.1 User Request (Q
0
)
This kind of request it is an informal query, which
represents the desired results from users. Thus, these
queries are used to describe the compositions goals
or to specific the input conditions for the service re-
trieval stage (very phase important into composition
process).
2.1.2 Definition 1 (Formal User Request Q)
Let Q
0
the informal user request expressed in natural
language before its analysis. A formal user request Q
is defined as a tuple Q = (userID, F, NF, C, λ) where
userID is the identifier of the user that perform the re-
quest, F is a non-empty finite set of elements that rep-
resent Functional Words, such that F = { f
1
, ..., f
n
}.
NF is a finite set of elements that represent Non-
Functional Words, such that NF = {n f
1
, ..., nf
m
}. F
and NF are distinct, F ∩ NF = Ø. C is a pair C =
(W, P) where W is a non-empty finite set of words
that denote Control Words (e.g., If, which, and, or,
etc.), such that W = {c
1
, ..., c
m
}, and P is a finite set
of elements that represent Punctuation Marks (,.;:).
λ = (F ×C) is the function that records the sentence
meaning/structure, which is helpful to generate the
logical mashup model (which is described in detail
in the following section).
Elements both F and C can represent Operators
O = {o}. A Operator is a basic unit both retrieval
and composition phases. Generally, o represents one
or more abstract services from a subset of existing real
services. NF is considered a Parameter P = {p} used
to refine the ranking of retrieved services.
2.1.3 Definition 2 (Folksonomy)
Folksonomies are being widely used in various tag-
ging applications of the Social Web. Folksonomies
reflect through tags the collective intelligence of a
crowd or a community (Wisdom of the Crowds)
in giving meaning to available resources (Power
Tags)(Helic et al., 2011).
Three main entities are identified in our pro-
posal: the user U = {u
1
, ..., u
n
}, the resources R =
{r
1
, ..., r
m
} and the tags T = {t
1
, ..., t
k
}. Users an-
notate the resources with tags, creating triple associa-
tions between the user, the resource and the tag. Thus,
the folksonomy can be defined by a set of annotations
A ⊆ U × R × T. A folksonomy can be considered as
an specific case of a Taxonomy τ, i.e., if τ is formed as
a folksonomy by people specifying one or more tags
t
j
to describe certain objects (in this case operators,
which represent web resources), the tags in τ are un-
related and τ is completely unstructured. Introducing
a taxonomy structure in τ, enhances query expressiv-
ity, and also helps keep tag-based descriptions suc-
cinct(Helic et al., 2011). In this sense, according to
above definitions, both Operators (Resources) o and
Parameters p can be described by a specific set of tags
d(o) ⊆ τ and d(p) ⊆ τ respectively, selected from the
taxonomy τ.
2.1.4 Definition 4 (Tag Query q)
In general, a tag query q ⊆ T ∈ τ selects a subset ψ
of an operator set O = {o} such that each operator in
the selected subset is described by all tags in q, taking
into account sub-tag relationships between tags, i.e.,
if a tag t
1
∈ τ is a sub-tag of t
2
∈ τ, denoted t
1
∼ t
2
.
Therefore, according to this, formally we have:
• ψ
f
(o) = {o ∈ O|∀t ∈ q
f
∃t
′
∈ d(o) : t
′
∼ t ∧∀ f
j
∈
F, ∃q
f
: q
f
⊆ T}
• ψ
c
(o) = { o ∈ O|∀t ∈ q
c
∃t
′
∈ d(o) : t
′
∼ t ∧∀c
j
∈
C, ∃q
c
: q
c
⊆ T}
• ψ
nf
(p) = {p ∈ P|∀t ∈ q
nf
∃t
′
∈ d(p) : t
′
∼ t ∧
∀n
f j
∈ NF, ∃q
nf
: q
nf
⊆ T}
Where: ψ
f
(o) and is an operator subset of all opera-
tor set O = {o} such that each operator in this subset
is described by all tags in q
f
(set of functional word
tags). Thus, for each f ∈ F there is a q
f
⊆ T. ψ
c
(o)
is an operator subset of all operator set O = {o} such
that each operator in this subset is described by all
tags in q
c
(set of control word tags). ). Thus, for each
c ∈ W there is a q
c
⊆ T. ψ
nf
(p) is an parameter subset
of all parameter set P = {p} such that each parame-
ter in this subset is described by all tags in q
nf
(set of
non-functional word tags). Thus, for each n
f
∈ NF
there is a q
nf
⊆ T.
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
804

2.2 NGM Model
A data mashup model can be expressed as a tuple
m = {userID, name, T, O, C, M, reputation}, where
userID is the identifier of user that perform the re-
quest, name is the unique name of the mashup, T is
the set of tags that describes it, O is the set of opera-
tors used in the mashup. C is the set of data flow con-
nectors ruling the propagation of data among opera-
tors, M is the set of data mappings of output attributes
to input parameters of connected operators, and repu-
tation counts how many times the mashup m has been
used (e.g., to compute rankings). Specifically:
2.2.1 Definition 5, Operators (O)
At a logical level operators O
l
are defined as a set
O
l
= {O
li
|O
li
= (name
i
, T
i
)} with name
i
being the
unique name of the operator o
li
and T
i
represents
a description based on tags of the o
li
(from the
some user). However, at an executable level, i.e.,
of composition patterns, which include sequence op-
erations, parallel operations, etc. O
p
= {O
pi
|O
pi
=
(In
i
, Out
i
, Op
i
)} is a non-empty set of operators,
where In
i
= {in
i0
, ..., in
ij
}, Out
i
= {out
i0
, ..., out
ik
}
and Op
i
= {Op
i0
, ..., Op
il
} are respectively the sets
of input, output, and operations of an operator op
i
.
Thus, the set of Operators O is defined as: O =
O
l
∪ O
p
. We distinguish three kinds of operators:
• Source operators, which fetch data from the web
or the local machine. They don’t have inputs, i.e.,
In
i
=
/
0.
• Typical operators, which consume data in input
and produce processed data in output. Therefore,
In
i
, Out
i
6=
/
0.
• Control operators, which are composition struc-
ture patterns: Sequential, AND-Split (Fork), XOR-
Split (Conditional), AND-Join (Merge) and XOR-
Join (Trigger)(Yu et al., 2007).
2.2.2 Definition 6, Data Flow Connectors (C)
Let’s C = {c
m
|c
m
∈ O×O : C ∩ O = ∅} the data flow
connectors that assign to each operator o
j
its prede-
cessor o
k
(where: j 6= k) in the data flow.
2.2.3 Definition 7, Data Mapping (M)
Let’s M the data mapping represents the set of
data mapping of the data flow from output param-
eters of an operator o
j
to input parameters of the
predecessor operator o
k
(where: j 6= k), as fol-
lows: M = {m
n
|m
n
∈ In× Out : In ∩ Out = ∅, In =
∪
i, j
in
i, j
, Out = ∪
i, j
out
ik
} In order to better understand
the formalisms defined above, the Figure 1 shows our
proposal for a Mashups’ meta-model, which is in-
deed very simple: only requires 13 concepts suffice to
model its composition features at an executable level
(abstractness).
Figure 1: NGM metamodel.
Given the described models of query Q' and
Mashup M we create the mashup m meets the user’s
request from the large number and variety of re-
sources on the currently Web in two ways: firts, we
generate a Logical Mashup Model (LMM) by analyz-
ing the user’s request in natural language and then,
we generate an Executable Mashup Model (EMM ab-
stractness) by matching LMM against our knowledge
base. Thus, the Algorithm on Figure 2, details this
strategy and summarizes the logic implemented by
the generation of EMM.
Figure 2: General EMM algorithm.
In line 4, we get the formal query Q from the
user’s request Q
o
in natural language by the NLA
function (Natural Language Analyzer)(Pedraza et al.,
2011). Then the GetLogicalMashupModel() func-
tion gets the LMM from the Q. In line 6, the
GetEmmFromRepository() function gets an EMM
abstractness, which has been previously generated by
the same user or other users of our platform. If the
WISHFULANDWISDOMAWARECOMPOSING-AFullUser-centricApproachtoCreateNGM
805

algorithm finds an exact or similar EMM, it is rec-
ommended to the user, avoiding the whole process of
composition. In the absence of an EMM that satis-
fies the user’s request, the EMM is composed based
on retrieved operators and the LMM obtained (be-
tween lines 10 and 22). Finally, the EMM generated
is stored in a Repository of abstractness EMM.
3 CONCLUSIONS AND FUTURE
WORK
The result of our research indicates that there is still
a lack approaches to provide a feasible solution for
end-users to mash the great diversity of existing re-
sources. In this paper, we proposed an hybrid inte-
gration of different types of available web resources.
We call this combination Next Generation Mashups
(NGM). To achieve this, we define a user-centric ap-
proach to create NGMs based on W2AC (Wishful and
Wisdom Aware Composing), a composition paradigm
that aims at determining what ordinary users really
want (Wishes) from a request in natural language, to
finally deliver them the best solution that meets their
needs (without requiring programming or technical
skills). To do this, we define two meta-models, one
to describe the user’s request and another to represent
the NGMs. Currently we have implemented the mod-
ule that generates the LMM described previously. The
next step of this work is to study and define new fea-
tures that extend the NGM meta-model.
ACKNOWLEDGEMENTS
The authors would like to thank Universidad del
Cauca, COLCIENCIAS and TelComp2.0 Project for
supporting the Research of the M.Sc. Student Luis
Javier Suarez.
REFERENCES
Agarwal, N., Galan, M., Liu, H., and Subramanya, S.
(2010). Wiscoll: Collective wisdom based blog clus-
tering. Inf. Sci., 180:39–61.
Casati, F. (2011). How end-user development will save
composition technologies from their continuing fail-
ures. In Proc. of the 3th international conference on
IS-EUD'11, pages 4–6, Berlin, Heidelberg.
Chowdhury, S. R., Daniel, F., and Casati, F. (2011). Ef-
ficient, interactive recommendation of mashup com-
position knowledge. In Kappel, G., Maamar, Z.,
and Motahari-Nezhad, H. R., editors, ICSOC, volume
7084 of LNCS, pages 374–388. Springer.
Chowdhury, S. R., Rodr´ıguez, C., Daniel, F., and Casati, F.
(2010). Wisdom-aware computing: on the interactive
recommendation of composition knowledge. In Pro-
ceedings of the ICSOC, pages 144–155, Berlin, Hei-
delberg. Springer-Verlag.
Dalal, N. (2008). Wisdom networks: Towards a wisdom-
based society. pages 11–18. Springer Berlin Heidel-
berg.
Dastani, M. and Steunebrink, B. (2009). Modularity in bdi-
based multi-agent programming languages. In Pro-
ceedings of the 2009 IEEE/WIC/ACM, WI-IAT ’09,
pages 581–584, Washington, DC, USA.
De Angeli, A., Battocchi, A., Chowdhury, S. R., Rodriguez,
C., Daniel, F., and Casati, F. (2011). End-user require-
ments for wisdom-aware eud. In Proceedings of the
3th IS-EUD'11, pages 245–250, Berlin, Heidelberg.
Springer-Verlag.
Helic, D., Strohmaier, M., Trattner, C., Muhr, M., and
Lerman, K. (2011). Pragmatic evaluation of folk-
sonomies. In Proceedings of the 20th WWW '11,
pages 417–426, New York, NY, USA. ACM.
Maaradji, A., Hacid, H., Skraba, R., and Vakali, A. (2011).
Social web mashups full completion via frequent se-
quence mining. In Proceedings of the SERVICES '11,
pages 9–16, Washington, DC, USA. IEEE Computer
Society.
Maries, I. and Scarlat, E. (2011). Enhancing the compu-
tational collective intelligence within communities of
practice using trust and reputation models. LNCS,
pages 74–95. Springer Berlin / Heidelberg.
Pedraza, C., Zuiga, J., Suarez, L. J., and Corrales, J. C.
(2011). Automatic service retrieval in converged envi-
ronments based on natural language request. In SER-
VICE COMPUTATION 2011, pages 52–56, Rome,
Italy.
Riabov, A. V., Boillet, E., Feblowitz, M. D., Liu, Z., and
Ranganathan, A. (2008). Wishful search: interactive
composition of data mashups. In Proceedings of the
17th WWW '08, pages 775–784, New York, NY, USA.
Szuba, T., Polanski, P., Schab, P., and Wielicki, P. (2011).
On efficiency of collective intelligence phenomena.
In Nguyen, N. T., editor, Transactions on computa-
tional collective intelligence III, chapter On efficiency
of collective intelligence phenomena, pages 50–73.
Springer-Verlag, Berlin, Heidelberg.
Yu, T., Zhang, Y., and Lin, K.-J. (2007). Efficient algo-
rithms for web services selection with end-to-end qos
constraints. ACM Trans. Web, 1:1–26.
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
806