A SURVEY ON WEB SERVICE DISCOVERING AND COMPOSITION

Elena del Val Noguera and Miguel Rebollo Pedruelo

Department of Information Systems and Computation

Technical University of Valencia, Camino de Vera s/n, Valencia, Spain

Keywords:

Web services, semantic web, intelligent agents, discovering, composition, planning, model checking.

Abstract:

This paper reviews the existing techniques used in the discovering and composing of services. The task of

selecting an adequate service can quickly grow tedious if all services that are listed under a certain description

have to be compared manually for the ﬁnal selection. And what is more, the ﬁnal selection does not only

depend on service parameters like executions costs or accuracy, but depends on the usefulness of objects or

information that service offers. This problem is present in open environments where entities like web services

or agents need to locate other entities to achieve cooperation, delegation or interoperation. For these reason

these two approaches, web services an agents have deal with these problem proposing an automated and

efﬁcient mechanism to determine a structural and semantic match descriptions between entities.

1 INTRODUCTION

The application of thesemantic web in the area of web

services has as aim a more intelligent web in which is

possible to achieve a more effective communication

among computers. It concentrates its efforts towards

web service semantic descriptions search. Descrip-

tive languages are not enough to describe complex re-

lationships between ontologies. For that reason, other

languages as OWL-S or WSMO have been proposed

with the aim that other machine can read these de-

scriptions and reason about how interact with the ser-

vices.

Automatic location of services can considerably

reduce the cost of making applications that work to-

gether and enable a more ﬂexible integration, where

providers are dynamically selected based on what

they provide and other non-functional properties. To

deal with these issues agent orientation is an ap-

propriate design paradigm to enforce automatic and

dynamic collaborations, especially in e-business and

complex environments.

In this paper we present a revision of the different

solutions that have contributed to deal with problem

of the service discovery. The rest of the paper is struc-

tured as follows. Section 2 gives a service discovery

description and it also presents the fundamentalstages

that are involve in the discovery process. Section 3

introduces the main approaches used in matchmaking

algorithms. Section 4 surveys the existing state of the

art in web service matchmaking algorithms. Section 5

the contribution of agent systems to service discovery

process and we close with a summary and concluding

remarks in section 6.

2 SERVICE DISCOVERY

Semantic service discovery consist of searching ser-

vice descriptions with exact or similar properties.

These properties, in most cases, are IOPE’s (Inputs,

Outputs, Preconditions and Effects). Basically, the

matchmaking process consist of bounding the number

of possible matches between offered and requested

services. This matchmaking process can be divided

in three stages:

Selection Process. A service request is receivedand

sent to a matchmaker. The matchmaker is responsible

of ﬁnding a suitable service or set of services accord-

ing with the requested description. The suitability de-

pends on the information that the algorithm considers.

Usually, the degree of similarity between service de-

scription and the request is used and it will depend

on the degree of similarity between input and output

parameters (IO’ s).

135

del Val Noguera E. and Rebollo Pedruelo M. (2008).

A SURVEY ON WEB SERVICE DISCOVERING AND COMPOSITION.

In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 135-142

DOI: 10.5220/0001524101350142

 SciTePress

Ranking. The set of possible service providers is

reﬁned according to additional information that the

client has deﬁned previously to choose the more suit-

able provider. The concept of non functional at-

tributes, in most of cases related to quality of ser-

vice (QoS), is introduced. These attributes are used

in ranking functions that gives a mark or punctuation

to each provider selected in the previous stage. Fur-

thermore, the user can deﬁne an threshold to ﬁlter the

obtained services.

Evaluation. When the results of the matchmak-

ing process are obtained, it is recommended evalu-

ate these results to identify possible modiﬁcations in

some matchmaking parameters. The most important

points to be evaluated are the quality of the results

and the system execution. Precision and recall are the

most commonly used measures.

3 SERVICE DISCOVERY

APPROACHES

At the beginning, all the approaches used syntactic

similarity to establish the degree of matching between

two service descriptions. With the arrival of semantic

web, service descriptions include data structures and

also relationships between other concepts, restrictions

and rules. At this situation, the proposed algorithms

are based on semantics. Furthermore, in some com-

plex environments, where negotiations and protocols

to interact between entities, it can be possible to es-

tablish discovery models based on them.

Service Discovery based on Keywords. Services

can be ﬁltered by doing a search based on keywords.

A typical keyword scenario is a search engine that re-

ceives keywords as a query and the engine engages

query keywords with service description keywords.

In (Bachlechner et al., 2006) there is a review of this

kind of discovery process.

Service Discovery based on Semantics. Keywords

do not use explicit semantic and, therefore, they do

not allow to make inferences to achieve better search

results. The use of vocabularies with a formal and

explicit semantic is considered as a second approach.

Ontologies give us a shared and explicit terminology

to describe web services and queries with a logic for-

malization that will allow the use of inference. We

found different approaches according to the service

description language used. Among them, the most

important ones use OWL-S, WSMO and SAWSDL

(based on WSDL-S) languages. These languages pro-

vide the answer to the main questions that arise when

a web service has to be described: What are the ser-

vice requirements from the users? What the service

provides to the users? How does the service work?

How the service can be used?.

Service Discovery based on Complex Discovery

Model and Negotiation/Contracting. It comprises

discovery models based on interaction protocols, for-

warding QoS and privacy requirements, negotiation

dialogs for reﬁning discovery and establishing ser-

vice requirements. This kind of service discovery is

also being used in multiagents systems (Caceres et al.,

2006) and in e-market environments.

4 SERVICE DISCOVERY BASED

ON SEMANTICS

There are many matchmaking proposals in the area of

web services and semantic web. In this section, we

review some of them paying attention to characteris-

tics such as the information that they use to deal with

the matchmaking process, if they take into account

service composition or cross ontologies or if they use

QoS information during the discovery process.

4.1 Iope’s Algorithms

In general, most of the existing matchmaking al-

gorithms use just inputs and outputs to determine

the matching degree between requests and advertise-

ments. But there are some algorithms that, apart from

that information, take into account preconditions and

postconditions (IOPE parameters).

The ﬁrst service discovery algorithm is based on

DAML-S (OWL-S predecessor) and uses IO’s. It was

developed by Paolucci et al. (Paolucci, 2002). This

approach uses the semantic of the Service Proﬁle and

UDDI registries to maintain the descriptions of the

services. The algorithm deals with the importance

in matchmaking classiﬁcation of the service outputs.

A matching between a service advertisement and a

service request consists of matching all the service

request outputs with those of the service advertise-

ment; and all the inputs of the service advertisement

with those of the service request. The degree of sim-

ilarity between service provider and server request

will depend on the degree of similarity between in-

put and output parameters (IO’ s) and it is reduced

generally to the minimal distance between them in

the taxonomic tree. The denomination of the degrees

varies according to literature (Abela and Montebello,

2002)(Lei and Horrocks, 2003)(Paolucci, 2002)(Con-

stantinescu and Faltings, 2002). Paolucci’s algorithm

WEBIST 2008 - International Conference on Web Information Systems and Technologies

136

is limited to discover simple services. It does not

consider the composition discovery nor the use of

different ontologies. Another lack is that the pro-

cess of matching does not consider parameters re-

lated with quality of service (QoS). From this algo-

rithm arose others that made some modiﬁcation to

deal with some of its deﬁciencies (Abela and Mon-

tebello, 2002)(Klusch et al., 2006)(Aversano et al.,

2004)(Cardoso and Sheth, 2002).

(Wolf-Tilo and Matthias, 2003) deﬁne a different

matchmaking algorithm. In this algorithm, they make

ﬁrst a search based on keywords and, once obtained

a list of results, the user could introduce IO’s param-

eters that must have the services and the values for

these parameters. Finally, a reasoner eliminates those

services that do not have these deﬁned parameters and

the resultants will be executed with the values that

were introduced by the client. The results of the ex-

ecutions are ordered following some constraints that

beneﬁt the client, as the quality of service.

(Klusch et al., 2006) present OWLS-MX, a hy-

brid matchmaking algorithm that computes the de-

gree of semantic matchmaking for a given pair of

service advertisement and request by successively ap-

plying ﬁve different ﬁlters: exact, plug-in, subsumes,

subsumed-by and nearest-neighbor. The ﬁrst three

are logic based only whereas the last two are hybrid

due to the required additional computation of syntac-

tic similarity values. The objective of hybrid seman-

tic web service matching is to improve semantic ser-

vice retrieval performance by appropriatelyexploiting

means of both crisp logic based and approximate se-

mantic matching.

In the matchmaking process is also important to

consider the global schema of execution, which is

given by the choreography. The task of selecting a

web service, that should play a role in a choreogra-

phy, implies verifying two things: the conformance

of the service to the speciﬁcation of a role of interest

(guarantees that the message exchange will produce

correct and accepted conversations), and that the use

of that service (allows the achievement of the goal).

In (Baldoni et al., 2007) is shown that performing

a match operation by operation does not preserve the

global goal. They also show how to overcome these

limits by exploiting the choreography deﬁnition. Ac-

tually, it is possible to extract from the choreography

some information that can be used to bias the match-

ing process so that the global goal will be preserved.

4.2 Composition

The algorithms above presented address the match-

making process. However, they are limited to dis-

cover a single service. In many situations, queries

that cannot be satisﬁed by a single service might be

frequently satisﬁed by composing several services.

(Aversano et al., 2004) display a discovery algo-

rithm that analyses DAML-S service proﬁle, takes as

objective the outputs of the user request and consid-

ers the possibility of reaching it with only one service.

If it is not possible, the method includes a backward-

chaining algorithm with the purpose of verifying the

possibility of ﬁnding a match for the user request by

means of a composition of several services. This al-

gorithm is also capable of performing a cross ontol-

ogy matching for service descriptions that use dif-

ferent ontologies. However, it crosses ontologies at

query time, hence severely affecting the efﬁciency of

the whole procedure.

The matching algorithms described until now are

based on DAML-S/OWL-S and use the service pro-

ﬁle. Analyzing web services only through their ser-

vice proﬁle (i.e., their IOs), can severely affects the

process of discovery of service aggregations that sat-

isfy a request. Indeed, the service proﬁle does not

describe the internal behavior of services, so in some

cases it does not provide valuable information needed

for composing services.

The ﬁrst discovery algorithm based on the anal-

ysis of the OWL-S Process Model was proposed by

(Bansal and Vidal, 2003). It stores advertisements

of services as tree structures corresponding to their

process models. The compound processes correspond

with intermediate nodes whereas the atomic processes

correspond with the leaves. The matchmaking algo-

rithm begins in the root of the tree of the advertise-

ment of the service and visits all the subtrees ﬁnish-

ing in the leaves. For each node, the algorithm veriﬁes

the compatibility between the IOs and the IOs of the

request.

SAM (Brogi et al., 2003) is an extension of the

Bansal algorithm that return, when a complete match-

making is not possible, a list of partial matchings (a

composition of subservices that can provide only cer-

tain requested outputs by the client). Besides, when

it does not ﬁnd any match, SAM is able to suggest to

the user additional inputs that can be enough to reach

complete match. The main lack of this algorithm is

that does not consider the use of different ontologies.

Another interesting point to take into account is

related with goals. Languages as WSMO, that con-

sider goals to achieve a composition. The work pre-

sented in (van Riemsdijk and Wirsing, 2007) points

out how goal-oriented techniques, which increase

ﬂexibility in handling failures, can be applied in the

context of service-oriented systems and speciﬁcally

in web services composition.

A SURVEY ON WEB SERVICE DISCOVERING AND COMPOSITION

137

4.3 Crossing Ontologies

Currently, individual users or user communities hope

to be able of making queries about interesting ser-

vices using descriptions that are expressed in terms

of their own ontologies, which do not have to ﬁt in

with the searched service descriptions. The above-

named algorithms do not address properly the prob-

lem of crossing ontologies.

(Cardoso and Sheth, 2002) propose an algorithm

that allows to manage multiple ontologies. The simi-

larity function used to compare concepts is based on

the ontology taxonomy. The algorithm tries to man-

age concepts that are not related using the concepts

properties. (Aversano et al., 2004) present a cross on-

tology matching that can cancel the problem of differ-

ent ontologies. Therefore, there is no need of classify

in a semantic domain the web service when creating

the service description.

(Pathak et al., 2005) propose an ontology mapping

during service discovery, such that terms and con-

cepts in the service requester’s ontologies are brought

into correspondence with the service provider’s on-

tologies. To do the mappings they use interoperation

constraints, i.e. a set of relationships that exist be-

tween elements from two different hierarchies.

(Brogi et al., 2006) present an extension of SAM

based on hypergraphs who allows to cross differ-

ent ontologies. The matchmaking system consists of

two main modules: the Hypergraph Builder and the

Query Solver. The Hypergraph Builder analyzes the

ontology-based descriptions of the registry-published

services in order to build a labeled directed hyper-

graph, which synthesizes all the data dependencies of

the advertised services. The vertexes of the hyper-

graph correspond to the concepts deﬁned in the on-

tologies employed by the analyzed service descrip-

tions, while the hyperedges represent relationships

among such concepts (subConceptOf, equivalentCon-

ceptOf and intra-service dependency). The Query

Solver explores the hypergraph by suitably consider-

ing the intraservice and inter-service dependencies to

address the discovery of (compositions of) services as

well as by considering the subConceptOf and equiva-

lentConceptOf relationships to cope with differenton-

tologies.

4.4 Hypergraphs

A hypergraph is a generalization of a graph, where

edges (hyperedges) can connect any number of ver-

texes. Formally, a hypergraph is a pair (V, E) where

V is a set of nodes or vertexes and E is a set of non-

empty subsets of V called hyperedges. While graph

edges are pairs of nodes, hyperedges joints arbitrary

sets of nodes.

(Yang et al., 2005) use arc-labeled and arc-

weighted trees to represent product/service require-

ments and offers. They propose a tree similarity al-

gorithm that traverses input trees top-down and then

computes their similarity bottom-up. During tree sim-

ilarity computation, when a subtree in T

is miss-

ing in tree T

(or viceversa), the algorithm compute

the simplicity of the missing subtree. The tree sim-

plicity measure takes into account the node degree

at each level, the depth of the leaf node and the arc

weights. This algorithm allows partial product de-

scriptions representations via subtrees missing.

(Hashemian and Mavaddat, 2005) use a speciﬁc

notation, called interface automata (state-base model)

in order to formally model web services. The infor-

mation that an interface automaton exposes is the IO

of a component and the temporal ordering of the ac-

tions it performs. This information can be extracted

form the OWL-S speciﬁcation of web services. Based

on the properties exposed by interface automaton of

each web service ws, three pieces of information are

stored in the repository: its set of inputs, its set of out-

puts and dependency information between IOs of the

web service. The repository is stored as a graph that

contains web services information. The nodes repre-

sents I/O and there is a directed edge form node v

to node v

if and only if there is a dependency be-

tween the input and the output. They solve the prob-

lem in two steps: (i) ﬁnding web services that can po-

tentially participate in the composition, and (ii) ﬁnd-

ing the composition setup based on the web services

found in the previous step.

4.5 Model Checking

Web services are composed online from pieces of

software created by different programmers. Individ-

ual services can be checked to ensure that they are

error free, but when new services are composed there

are no means to check whether the composed service

fulﬁls its purpose. Some formal methods, as model

checking, has been proposed to verify the correctness

of complex services. But current languages are semi-

formal, so the correctness of the composition depends

on the cleverness of the designer. To use formal mod-

els requires translations from the languages used to

describe WS into more formal ones.

(Gao et al., 2006) translate web services speciﬁed

in BPEL4WS into pi-calculus, which is nearer to pro-

gramming languages than ﬁnite automata or tempo-

ral logics. Nevertheless, this formal description is

not soundness and some manual translation is still

WEBIST 2008 - International Conference on Web Information Systems and Technologies

138

needed. The model checking is used with two pur-

poses: (i) to check if services satisfy customer’s de-

mands and and designer’s speciﬁcations, and (ii) to

check if orchestration satisﬁes liveness, safety, fair-

ness and reachability. Different methods are used:

bisimulation to verify the speciﬁcation, mu-calculus

to check properties as safety or reachability and pi-

calculus to eliminate ill behaviors.

Nakajima claims that to verify a composite web

service prior to its execution may be mandatory

(Nakajima, 2002). First, translates a WSFL descrip-

tion into Promela, the speciﬁcation language for SPIN

model checker. Furthermore, additional properties are

expressed in LTL to be added to the model checking

process. The veriﬁcation process detects reachability,

deadlock freedom and speciﬁc user properties.

Planning as model checking (Giunchiglia and

Traverso, 1999) is a method of solving planning prob-

lems modeling them as model checking problems.

This solution is based on transition systems, but web

services are a message passing paradigm, so some

special considerations have to be made. (Yu and

Reiff-Marganiec, 2006) make a formulation of the so-

lution by modifying the strong cycle planning algo-

rithm, which guarantees that all paths reach a solu-

tion and they are fair. A four-phased algorithm is pro-

posed. First, the planning goal and the initial knowl-

edge is speciﬁed. After that, automatically selects

from the repository relevant web services to build the

plan. In third place, the algorithm search for plans.

Finally, a physical composition step allows clients to

choose the better plan, generates a executable plan

speciﬁed in BPEL and monitors its execution, re-

planning when a failure is detected.

(Walton, 2004) uses model checking to validate

the correctness of communication protocols between

agents in an platform that integrates agents and web

services. The services are described in WSDL. Com-

plex interactions among the entities that offer services

are represented by the protocols, who are speciﬁed

in a directly executable language called MAP. As the

Nakajima’s algorithm, the speciﬁcation is translated

into Promela language. This one provides a complete

automatic translation that allows non-expert to vali-

date their services.

The main problem in all these approaches is the

complexity of the state space. All of them make dif-

ferent simpliﬁcations to the problem to be capable of

managing the validation process by limiting the num-

ber of services (or agents) and the length of the mes-

sage interchanging mainly. Moreover, designer’s in-

tervention is often needed to translate service descrip-

tions into formal languages for model checking.

4.6 Non-Functional Parameters

In some algorithms, the service selection process

is based in non-functional parameters and, in other

cases, the algorithms reﬁne the set of candidate ser-

vice providers based on user-speciﬁed non-functional

attributes, namely Quality of Service (QoS). These

factors and domain speciﬁc characteristics affect on

the service selection.

According to (J.Radatz and Sloman, 1988) the

Quality of Service is a set of non-functional attributes

that may impact the service quality offered by a web

service. The main problem of this kind of information

is that we cannot trust the QoS characteristics pub-

lished by provider.

(Pathak et al., 2005) establish a categorization

in two groups: domain-dependent and domain-

independent attributes. The domain-independent at-

tributes represent those QoS services characteristics

which are not speciﬁc to any particular service (for in-

stance, scalability or availability). On the other hand,

domain-dependent attributes capture those QoS prop-

erties which are speciﬁc to a particular domain and

most of the times are dynamic and depends on the in-

stant in which the service is executed.

To deal with information reliability, service repu-

tation is an specially interesting property that could be

regarded as a measure that accumulates a user opinion

about QoS in general. (Kalepu et al., 2004) address

this problem.

(Pathak et al., 2005) propose a taxonomy for

the non-functional attributes which provide a better

model for capturing various domain-dependent and

domain-independent QoS attributes of the services.

These attributes allow users to dynamically select ser-

vices based on their non-functional aspects. This

work also introduce the notion of personalized rank-

ing criteria, which enhances the traditional ranking

approach, primarily based on the degree of match.

Furthermore, in this work a kind of ontology mapping

is presented .

Following this trend, (Kokash, 2005) present an

approach based on the application of a distributed rec-

ommendation system to provide QoS information and

on testing of retrieval methods on service speciﬁca-

tions.

(Aversano et al., 2004) base the searching process

on syntactic information and on service quality met-

rics and semantics to increase the precision of the dis-

covery process. To take into consideration the QoS in

the discovery process, the customer of the service will

assign a weight to every attribute. For each selected

service, a weighted sum is performed among all the

attributes an the ﬁnal value represents the quality of

A SURVEY ON WEB SERVICE DISCOVERING AND COMPOSITION

139

the service.

4.7 Efﬁcient Matching

Due to the complexity of the underlying semantic rea-

soning, matching semantic web service capabilities is

a heavy process. Furthermore, the matchmaking pro-

cess could be intractable when the number of avail-

able services gets large. (Mokhtar et al., 2006) de-

scribe a solution towards the efﬁcient matching of se-

mantic service capabilities. This approach combines

optimizations of the discovery process at reasoning

and matching levels. Towards the optimization of

the discovery process at reasoning level, they use the

solution proposed by (Constantinescu and Faltings,

2002) for encoding concept hierarchies, represented

by hierarchies of concepts, using intervals. These hi-

erarchies represent the subsumption relationships be-

tween all the concepts in the ontologies used in the

directory. The main idea is that any concept is associ-

ated with an interval. Under the assumption that ser-

vice advertisements and service requests already con-

tain the codes corresponding to the concepts that they

involve, semantic service reasoning reduces to a nu-

meric comparison of codes.

(Srinivasan et al., 2004) present an approach to

optimize service discovery in a UDDI registry, aug-

mented with OWL-S for the description of seman-

tic web services. This approach is based on the fact

that the publishing phase is not a time critical task.

Therefore, the authors propose to exploit this phase to

pre-compute and store information about the incom-

ing services. This proposal increases the time spent

for publishing service advertisements, but it consider-

ably reduces the time spent to answer a user request

compared to approaches based on on-line reasoning.

The majority of the algorithms presented in the

area of web services faces the discovery problem us-

ing the IO’s parameters and trying to solve problems

such as service composition or cross ontologies and

also in some of them non-functional parameters are

taken into consideration. But they do not explote all

the possibilities that semantic offers.

5 AGENTS AND DISCOVERY

PROCESS

Nowadays, service oriented computing (SOC) brings

additional considerations, such as the necessity of

modeling autonomous and heterogeneous compo-

nents in uncertain and dynamic environment. Such

components must be autonomously reactive and

Table 1: Service Discovery Algorithms.

Alg. Lang. IO’S/ C. Cross QoS

IOPE’S Ont.

Paolucci02 DAML-S IO’s

Abela02 DAML-S IO’s

Constant.02 DAML-S IO’s

√

Cardoso02 DAML-S IO’s

√ √

Lei03 DAML-S IO’s

Wolf03 IO’s Soft

constr.

Bansal03 DAML-S IO’s

√

Brogi03 OWL-S IO’s

√

Benata.03 DAML-S IO’s

√

Aversa.04 DAML-S IO’s

√ √ √

PE’s?

Sriniv.04 OWL-S IO’s

√

KluS05 OWL-S IO’s

√

PE’s?

Pathak05 OWL-S IOPE’s

√ √

Hashem.05 OWL-S IO’s

√

Yang05 OWL-S IO’s

√

Kokash05 WSDL IO’s

√

Klusch06 OWL-S IO’s

√

Brogi06 OWL-S IO’s

√ √

Mokhtar06 DAML-S IO’s

√

proactive yet able to interact ﬂexibly with other com-

ponents and environments. Agent orientation is an

appropriate design paradigm to enforce automatic

and dynamic collaborations, especially for systems

with complex and distributed transactions. Agent

paradigm has technical advantages in software con-

struction, legacy systems integration, transaction-

oriented composition and semantics-based interac-

tion. For this reason, many ideas from the research in

multiagent systems could be used in service-oriented

computing approaches. Key MAS concepts are re-

ﬂected directly in SOC with ontologies, process mod-

els, choreography, directories and facilitators, service

level agreements and quality of service measures.

Agent paradigm has presented several ideas that

can be taking into consideration for web service dis-

covery algorithms. (Caceres et al., 2006), presents

an approach that complement the existing methods

by considering the types of interactions and roles

that services can be used in. Other approaches (Sen-

soy et al., 2007) use the objective experience data of

agents in order to evaluate its expectation from a ser-

vice provider and makes decisions using its own cri-

teria and mental state. Basically, service consumers

collect previous experiences form other service con-

sumers with similar demand and make decisions us-

ing different methods. By simply sharing their expe-

riences, service consumers lead to the emergence of a

consumer society in which increase the overall satis-

faction.

WEBIST 2008 - International Conference on Web Information Systems and Technologies

140

In some MAS, brokers are used as the mecha-

nism of discovery and synchronization mechanisms

among autonomous agents. (Sycara et al., 2004)

present a brokering protocol which consist on two

complex reasoning tasks: discovery and mediation.

Discovery task is divided in two different reasoning

tasks: abstraction from the query to the provider’s re-

quired capabilities and comparison and matching of

required capabilities with providers capabilities. Me-

diation task requires the broker to transform the orig-

inal query into one that can send to the provider.

Sometimes, matchmaking process is considered

as a query to a static set of available options. This

view is to simple. We can consider matching as a typ-

ical example of symmetry and iterative nature (Heep,

2006). Symmetric because the visibility of character-

istics might equally depend on whether the other party

meets speciﬁc criteria. Iterative because we learn the

option space by analyzing our initial query’s result

set, and might restrict or weaken our requirements

and preferences in response. Negotiation protocols

are another mechanism used in MAS. The participant

agents negotiate about the properties of the services

they request and provide to bind agreements and con-

tracts with each other. In (Dang and Hungs, 2006)

is presented a protocol that supports many-to-many

negotiation in which several agents negotiating with

many other agents simultaneously using colored petri

nets.

6 CONCLUSIONS

After the review process, some common weakness

have been detected. Most of them they do not exploit

all the data provided in the service proﬁle. The ma-

jority of the algorithms use IO’s but forget the precon-

ditions and postconditions and the capability to make

inference about this data to obtain new information

that could be useful to eliminate false suitable ser-

vices. Moreover, only a few take into account infor-

mation related with QoS and theyuse this information

just for ranking, they are not used to select services

or reduce the search space with this data. Further-

more, they not emphasize in other aspects that could

give them more ﬂexibility in the discovery process

such as allowing partial matching or fuzzy data. In

general, web services nor have internal state and nei-

ther awareness of changes in its environment and their

algorithms do not consider temporal constraints over

services.

Current web services technologies have not ex-

ploited sufﬁcient semantics and approaches to dy-

namic service-oriented operations in open environ-

ments. Agents present an extended proposals that pro-

vide a more ﬂexible and efﬁcient matchmaking pro-

cess. Some of the proposals use algorithms very simi-

lar to web services but they add use the characteristics

of the agents to achieve a more efﬁcient and ﬂexible

matching. We have seen in this article that the use of

roles, interaction patterns, trust or negotiation. Agents

solve problems present in web services such as in-

ternal state and communication. Agents have aware-

ness of theirs internal states and for communication

use asynchronous message interchange which allows

to stablish conversations. There are some open issues

such as consider QoS in matchmaking process and,

as in web services, take into account temporal con-

straints.

To conclude, after analyse the proposals from

these two worlds, the highest beneﬁt would be

achieved combining the advantages of each world.

The problem of service discovery is faced by the web

services using semantic languages, taking into ac-

count service composition and other points as cross-

ing ontologies or QoS. But these items provide a func-

tional matching that do not consider other informa-

tion, more subjective or in a higher level of abstrac-

tion, but that can be also important to provide a more

accurate service discovery that ﬁts better with the user

request. To achieve that, we have to consider no only

the information present in the current descriptions of

web services. We have also to consider the kind of in-

teractions between the entities, the type of role that is

necessary to interact with or the reputation of the in-

volved providers. These characteristics are useful to

limit the possible candidates to be a suitable provider

and to make easier to ﬁnd what you want. In some

web services proposals, these ideas are being taken

into account, but it is difﬁcult task due to the differ-

ence between web services and agents, although grad-

ually the difference between them is becoming lower.

REFERENCES

Abela, C. and Montebello, M. (2002). Daml enabled web

services and agents in the semantic web. In WS–

RSD’02, Germany.

Aversano, L., Canfora, G., and Ciampi, A. (2004). An al-

gorithm for web service discovery through their com-

position. In IEEE International Conference on Web

Services (ICWS04).

Bachlechner, D., Siorpaes, K., Lausen, H., and Fensel, D.

(2006). Web Service Discovery - A Reality Check.

Baldoni, Baroglio, Martelli, Patti, and Schifanella (2007).

Service selection by choreography-driven matching.

In ECOWS Workshop on Emerging Web Services

Technology.

A SURVEY ON WEB SERVICE DISCOVERING AND COMPOSITION

141

Bansal, S. and Vidal, J. (2003). Matchmaking of web ser-

vices based on the daml-s service model. In Second

International Joint Conference on Autonomous Agents

(AAMAS03). T. Sandholm and M. Yokoo.

Brogi, A., Corﬁni, S., Aldana, J., and Navas, I. (2006).

Automated discovery of compositions of services de-

scribed with separate ontologies.

Brogi, A., Corﬁni, S., and Popescu, R. (2003).

Composition-oriented service discovery. Proceed-

ings of 5th International Conference on Autonomous

Agents and Multi-Agent Systems (AAMAS).

Caceres, C., Fernndez, A., Ossowski, S., and Vasirani, M.

(2006). Role-based service description and discov-

ery. In International Joint Conference on Autonomous

Agents and Multi-Agent Systems.

Cardoso, J. and Sheth, A. (2002). Semantic e-workﬂow

composition. Journal of Intelligent Information Sys-

tems (JIIS).

Constantinescu, I. and Faltings, B. (2002). Efﬁcient match-

making and directory services. Technical Report No

IC/2002/77.

Dang, J. and Hungs, M. (2006). Concurrent multiple-issue

negotiation for internet-based services. In IEEE Inter-

net Computing, number Vol.10 - 6, pages 42–49.

Gao, C., Liu, R., Song, Y., and Chen, H. (2006). A model

checking tool embedded into services composition en-

vironment. In GCC ’06: Proceedings of the Fifth

International Conference on Grid and Cooperative

Computing (GCC’06), pages 355–362, Washington,

DC, USA. IEEE Computer Society.

Giunchiglia, F. and Traverso, P. (1999). Planning as model

checking. In ECP, pages 1–20.

Hashemian, S. and Mavaddat, F. (2005). A graph-based

approach to web services composition. In IEEE Com-

puter Society.

Heep, M. (2006). Semantic web and semantic web services.

In IEEE Internet Computing, number Vol.10 - 2, pages

85– 88.

J.Radatz and Sloman, M. (1988). A standard dictionary for

computer terminology: Proyect 610. IEEE Computer.

Kalepu, S., Krishnaswamy, S., and Loke, S. (2004). Reputa-

tion = f(user ranking, compliance, verity). In Proceed-

ings of the IEEE International Conference on Web

Services.

Klusch, M., Fries, B., and Sycara, K. (2006). Automated

semantic web service discovery with owls-mx. In

Proceedings of 5th International Conference on Au-

tonomous Agents and Multi-Agent Systems (AAMAS),

Hakodate, Japan.

Kokash, N. (2005). Web service discovery with implicit

qos ﬁltering. In Proceedings of the IBM PhD Student

Symposium.

Lei, L. and Horrocks, I. (2003). A software framework

for matchmaking based on semantic web technol-

ogy. In Twelfth International World Wide Conference

(WWW2003), Germany.

Mokhtar, S., Kaul, A., Georgantas, N., and Issarny, V.

(2006). Towards efﬁcient matching of semantic web

service capabilities. In Proc. of WS-MaTe06.

Nakajima, S. (2002). Model-checking veriﬁcation for re-

liable web service. In OOPSLA 2002 Workshop on

Object-Oriented Web Services, Seattle, Washington.

Paolucci, M. (2002). Semantic matching of web services

capabilities. In The First International Semantic Web

Conference.

Pathak, J., Koul, N., Caragea, D., and Honavar, V. (2005).

A framework for semantic web service discovery. In

WIDM’05, Germany.

Sensoy, M., Pembe, C., Zirtiloglu, H., Yolum, P., and Bener,

A. (2007). Experience-based service provider selec-

tion in agent-mediated e-comerce. In Engineering Ap-

plications of Artiﬁcial Intelligence, number 3, pages

325–335.

Srinivasan, N., Paolucci, M., and Sycara, K. (2004). Adding

owl-s to uddi implementation and throughput. In In

Workshop on Semantic Web Service and Web Process

Composition.

Sycara, K., Paolucci, M., Soudry, J., and Srinivasan, N.

(2004). Dynamic discovery and coordination of agent-

based semantic web services. In IEEE Internet Com-

puting, number Vol.8 - 3, pages 66–73.

van Riemsdijk, M. B. and Wirsing, M. (2007). Goal-

oriented and procedural service orchestration - a for-

mal comparison. In In MALLOW-AWESOME’007.

Walton, C. (2004). Model checking multi-agent web ser-

vices. In Proceedings of the 2004 Spring Symposium

on Semantic Web Services, Stanford, CA, USA.

Wolf-Tilo, B. and Matthias, W. (2003). Towards personal-

ized selection of web services. In The Twelfth Inter-

national WWW Conference en Budapest.

Yang, L., Sarker, B. K., Bhavsar, V. C., and Boley, H.

(2005). A weighted-tree simplicity algorithm for sim-

ilarity matching of partial product descriptions. In In

Proceedings of ISCA 14th International Conference

onIntelligent and Adaptive Systemsand Software En-

gineering, Toronto.

Yu, H. Q. and Reiff-Marganiec, S. (2006). Semantic web

services composition via planning as model checking.

Tech. Report CS-06-003, University of Leicester.

WEBIST 2008 - International Conference on Web Information Systems and Technologies

142