HAL Id: hal-00656155
https://hal.archives-ouvertes.fr/hal-00656155
Submitted on 3 Jan 2012
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
A Cooperative Answering Approach to Fuzzy
Preferences Queries in Service Discovery
Katia Abbaci, Fernando Lemos, Allel Hadjali, Daniela Grigori, Ludovic
Lietard, Daniel Rocacher, Mokrane Bouzeghoub
To cite this version:
Katia Abbaci, Fernando Lemos, Allel Hadjali, Daniela Grigori, Ludovic Lietard, et al.. A Cooperative
Answering Approach to Fuzzy Preferences Queries in Service Discovery. Flexible Query Answering
Systems, 2011, Belgium. pp.318-329. �hal-00656155�
A Cooperative Answering Approach to Fuzzy
Preferences Queries in Service Discovery
Katia Abbaci
1
, Fernando Lemos
2
, Allel Hadjali
1
, Daniela Grigori
2
,
Ludovic Liétard
3
, Daniel Rocacher
1
, and Mokrane Bouzeghoub
2
1
IRISA/ENSSAT, Rue de Kérampont BP 80518 Lannion, France
{katia.abbaci,allel.hadjali,daniel.rocacher}@enssat.fr
2
PRiSM Lab, 45 Av. des États Unis 78000 Versailles, France
{fernando.lemos,daniela.grigori,
mokrane.bouzeghoub}@prism.uvsq.fr
3
IRISA/IUT, Rue Edouard Branly BP 30219 Lannion, France
ludovic.lietard@univ-rennes1.fr
Abstract. In this paper, we propose a novel approach for service retrieval that
takes into account the service behavior (described as process model) and relies
both on preference satisfiability and structural similarity. User query and tar-
get process models are represented as annotated graphs, where user preferences
on QoS (Quality of Service) attributes (such as response time, availability and
throughput) are modelled by means of fuzzy sets. To avoid empty results, a flex-
ible evaluation method based on fuzzy linguistic quantifiers (such as almost all)
is introduced. The retrieved results are easily interpreted by the end users thanks
to the clear semantics conveyed by that method. Finally, two families of ranking
methods are discussed.
Keyw ords: Cooperative answering, service retrieval, quality of services, fuzzy
preferences, linguistic quantifiers.
1 Introduction
Nowadays, an increasing number of companies and organizations are moving towards a
service-oriented and model-driven architectures for offering their services on the Web.
Searching a specific service within service repositories becomes a critical issue for the
success of these architectures. This issue has recently received much attention and many
approaches have been proposed [8,2,5]. Most of these approaches are based on the
matchmaking of process inputs/outputs [8], service behavior [2] or ontological knowl-
edge [5]. Unfortunately, these approaches often result in a large number of services
offering similar functionalities and behavior. One way to discriminate between such
similar services is to consider non-functional requirements such as QoS (Quality of
Service) (e.g., response time, throughput, availability and reliability). A recent trend
towards quality-aware approaches has been initiated [13,1,18], but remains limited and
not satisfactory for generic process model discovery.
On the other hand, several service discovery approaches based on fuzzy set theory
have been proposed. For instance, in [11] the authors treat the web service selection
H. Christiansen et al. (Eds.): FQAS 2011, LNAI 7022, pp. 318–329, 2011.
c
Springer-Verlag Berlin Heidelberg 2011
A Cooperative Answering Approach to Fuzzy Preferences Queries 319
for composition as a fuzzy constraint satisfiability problem. They assign to each QoS
criterion five fuzzy sets (such as poorly acceptable, almost acceptable and acceptable)
describing its constraint levels. In [15], QoS based service selection is modelled as a
fuzzy multiple criteria decision making problem. Linguistic expressions are used to
evaluate and to express the weights of importance of QoS criteria. Hafeez et al. [6]
present a service selection mechanism allowing the service broker to intelligently se-
lect a set of available services from a user query with imprecise constraints defined by
fuzzy sets. The query evaluation is based on the aggregation of the obtained degrees
over constraints. ¸Sora et al. [1] propose an approach in which they automatically gen-
erate fuzzy rules from user preferences and rank the candidate services using a fuzzy
inference process.
The above fuzzy approaches only consider the preference satisfiability and ignore
the structural similarity of complex web services. Moreover, these works deal only with
services as black boxes, i.e., the service behavior level is not investigated. Our goal is to
go further these approaches into a unique integrated approach dealing with functional
and non-functional requirements and behavior specification in service retrieval.
Starting from the work done in [9], we propose a cooperative approach for handling
users process queries where both behavior specification and QoS preferences are spec-
ified inside these queries. User preferences on QoS properties are modelled by means
of fuzzy sets as they are more suitable to the interpretation of linguistic terms (such as
high or fast) that constitutes a convenient way for users to express their preferences. To
avoid empty answers for a given query, a flexible evaluation strategy based on fuzzy
linguistic quantifiers is introduced.
The remainder of this paper is organized as follows. Section 2 provides some basic
background. In Section 3, modelling fuzzy preferences and their evaluation are ad-
dressed. Section 4 presents our interpretation of process models similarity based on
linguistic quantifiers. In Section 5, service ranking methods are discussed. Section 6
proposes an illustrative example and finally Section 7 concludes the paper.
2 Background
In this section, we provide some basic definitions within the scope of web service se-
lection with preferences, and a short recall on fuzzy sets.
2.1 Fuzzy Sets
A fuzzy set F [17] on the universe X is described by a membership function μ
F
:
X → [0, 1] , where μ
F
(x) represents the membership degree of x in F .Theset
{x ∈ F |μ
F
(x) > 0} (resp. {x ∈ F |μ
F
(x)=1}) represents the support (resp. core)
of F . In practice, the membership function associated to F is often represented by a
trapezoid (α, β, ϕ, ψ)
1
,where[α, ψ] (resp. [β,ϕ]) is its support (resp. core).
A Fuzzy set-based approach to preferences queries proposed in [3] relies on the use
of fuzzy set membership functions that describe the preference profiles of the user on
each attribute domain involved in the query. This is especially convenient and suitable
1
In our case, the quadruplet (α, β, ϕ, ψ) is user-defined to ensure the subjectivity property.
320 K. Abbaci et al.
when dealing with numerical domains, where a continuum of values is to be interfaced
for each domain with satisfiability degrees in the unit interval scale. Then individual
satisfiability degrees associated with elementary conditions are combined (commen-
surability assumption holds thanks to the membership functions) using a panoply of
fuzzy set connectives, which may go beyond conjunctive and disjunctive aggregations
(by possibly involving fuzzy quantifiers, if only the satisfiability of the most of the
elementary conditions in a query is required).
2.2 Preferences in Process Model Specification
Many languages are currently available to describe service process models, e.g., OWL-
S [12]. They represent a process model as a set of primitive activities combined using
control flow structures. Then, these languages can be abstracted as a direct graph G =
(V,E), where the vertices represent activities or control flow nodes, while the edges
represent the flow of execution. In this work, services are specified as graphs annotated
with QoS properties and user queries are specified as graphs annotated with preferences.
Figure 1 shows an example of a user query annotated with preferences. The example
presents a global preference indicating user prefers services providing RSA encryption.
Some activity preferences are also defined for activities A and B involving reliability,
response time and cost. Figure 2 shows an example of a process model annotated with
QoS attributes. The example presents a global annotation indicating the security of the
process model and activity annotations indicating the response time, reliability and cost
of some activities. In what follows, we present the formal definitions of our model.
start
AND
C
AND
end
ଶ
ǣ݉ܽݔݎ݈ܾ݈݁݅ܽ݅݅ݐݕ
ଷ
ǣܽݎݑ݊݀ݎ݁ݏܶ݅݉݁ǡʹͲ݉ݏ
ସ
ǣܾ݁ݐݓ݁݁݊ܿݏݐǡͳͲǡʹͲ
ହ
ǣƬ
ଶ
ǡ
ଷ
ǣƬ
ଶ
ǡ
ସ
ǣ۪
ଷ
ǡ
ସ
ଵ
ǣ݈݅݇݁ݏݏ݁ܿݑݎ݅ݐݕǡܴܵܣ
A
B
଼
ǣ݉ܽݔݎ݈ܾ݈݁݅ܽ݅݅ݐݕ
ଽ
ǣ݉݅݊ݎ݁ݏܶ݅݉݁
ଵ
ǣƬ
଼
ǡ
ଽ
ଵଵ
ǣ݉ܽݔݎ݈ܾ݈݁݅ܽ݅݅ݐݕ
ଵଶ
ǣܾ݁ݐݓ݁݁݊ܿݏݐǡͺǡͳͷ
ଵଷ
ǣܽݎݑ݊݀ݎ݁ݏܶ݅݉݁ǡͳͷ݉ݏ
ଵସ
ǣƬ
ଵଶ
ǡ
ଵଷ
Fig. 1. Query Graph q
1
start
end
ܽ
ହ
ǣሺݎ݁ݏܶ݅݉݁ǡʹͷ݉ݏሻ
ܽ
ǣሺݎ݈ܾ݈݁݅ܽ݅݅ݐݕǡͻͲΨሻ
AND
AND
B'
ܽ
଼
ǣ ݎ݁ݏܶ݅݉݁ǡ͵ͷ݉ݏ
ܽ
ଽ
ǣሺܿݏݐǡሻ
ܽ
ଵ
ǣሺݎ݈ܾ݈݁݅ܽ݅݅ݐݕǡͻͷΨሻ
D'
ܽ
ǣሺݎ݁ݏܶ݅݉݁ǡʹͷ݉ݏሻ
C'
ܽ
ଵ
ǣ ݏ݁ܿݑݎ݅ݐݕǡܴܵܣ
ܽ
ଶ
ǣ ݎ݁ݏܶ݅݉݁ǡͳͷ݉ݏ
ܽ
ଷ
ǣ ݎ݈ܾ݈݁݅ܽ݅݅ݐݕǡͺͲΨ
ܽ
ସ
ǣ ܿݏݐǡʹͷΨ
A'
Fig. 2. Target Graph t
1
Definition 1. An annotation is a pair (m, r),wherem is a QoS attribute obtained from
an ontology O and r is a value for m
2
. It can be specified over a process model graph
(global annotation) or over an atomic activity (activity annotation).
2
We abstract from the different units in which a value can be described.
A Cooperative Answering Approach to Fuzzy Preferences Queries 321
Definition 2. A preference is an expression that represents a desire of the user over the
QoS attributes of a process model or activity. It can be of one of the following forms
3
:
– atomic preferences:
• around ( m, r
desired
,μ
around
): for attribute m, this expression favors the value
r
desired
; otherwise, it favors the values close to r
desired
. μ
around
evaluates the
degree to which a value r satisfies r
desired
;
• between (m, r
low
,r
up
,μ
between
): for attribute m, this expression favors the
values inside the interval [r
low
,r
up
]; otherwise, it favors the values close to the
limits. μ
between
evaluates the degree to which a value r satisfies the interval
[r
low
,r
up
];
• max (m, μ
max
): for attribute m, this expression favors the highest value; oth-
erwise, the closest value to the maximum is favored, as example: the maximum
of reliability or availability is equal by default to 100%. μ
max
evaluates the
degree to which a value r satisfies the highest value of m;
• min (m, μ
min
): for attribute m, this expression favors the lowest value; oth-
erwise, the closest value to the minimum is favored, as example: the minimum
of response time or cost is equal by default to 0. μ
min
evaluates the degree to
which a value r satisfies the lowest value of m;
• likes (m, r
desired
): for attribute m, this expression favors the value r
desired
;
otherwise, any other value is accepted to some extent;
• dislikes (m, r
undesired
): for attribute m, this expression favors the values that
are not equal to r
undesired
;otherwise,r
undesired
is accepted to some extent;
– complex p references:
• Pareto preference ⊗ (p
i
,p
j
): this expression states that the two preference ex-
pressions p
i
and p
j
are equally important;
• prioritized preference &(p
i
,p
j
): this expression states that the preference p
i
is more important than the preference p
j
.
A preference can be specified over a process model graph (global preference) or over
an atomic activity (activity preference).
In [9], this set of preferences has been used to develop a service selection approach
based on QoS where preference satisfiability is computed using to a unique distance
function for all numerical preferences. This way of doing does not take into account the
fact that preferences are context and user-dependent and assumes no commensurability
when combining individual satisfiability degrees.
3 A Fuzzy Model to Evaluate Preferences
In this section, we introduce a fuzzy set-based approach to handle the above set of pref-
erences involved in the annotated graph associated with the user query. In particular, we
propose a metric, called satisfiability degree (δ), that measures how well the annotations
of a target process model satisfy the preferences present in the query.
3
Based on a subset of preference operators of the model by [7] that leads to a partial order.
