PERSONALIZED WEB SERVICE DISCOVERY

Sanaullah Nazir, Brahmananda Sapkota and Tomas Vitvar

Digital Enterprise Research Institute, National University of Ireland, Galway, IDA Business Park, Galway, Ireland

Keywords:

Semantic Web, Semantic Web Services, Ontology, Semantic Web service discovery, e-Government, Personal-

ization.

Abstract:

Most of the existing work in web service discovery reﬂect providers perspective. These discovery techniques

improve service descriptions so that the service clearly describes what it offers. This development has led to the

use of semantic web technology for web service discovery. A lot of work has gone into improving web service

descriptions to support improved web service discovery. In this paper we extend the existing approaches to

semantic web service discovery and introduce personalization to it. We also show how personalization affects

discovery. We provide a model to represent a user proﬁle, and its usage for improving the accuracy of web

service discovery.

1 INTRODUCTION

Advancement in web service discovery is mostly

centred around the service provider. Use of

UDDI (Clement et al., 2004) has been amongst the

initial efforts for improving web service discovery.

The inability of web service description language

WSDL to describe what the service provides led to

the use of semantic technology, involving DAML-

S (Paolucci et al., 2002), OWL-S (Srinivasan et al.,

2004), WSMO (Keller et al., 2004) etc. Most of the

currently used discovery approaches has been geared

towards identifying services which meet the user’s

requirements. There might be some web services,

which provide the same output, but the user of the

service could be different.

Context based service discovery (Hai et al.,

2005), (Maamar et al., 2006) is a type of service

discovery approaches which concentrate on the user’s

perspective. But the direction of context based dis-

covery has been dependent on environmental param-

eters in which the user is searching for a service.

Another discovery approach is QoS based service

discovery (Vu et al., 2005), (Makripoulias et al.,

2006). In this approach some numeric parameters are

set for some characteristics of a service. These val-

ues should match with the requirements of the service

user, e.g the required shipment time of the product

should match with the shipment time provided by the

service.

Here it is important to understand that the discov-

ery approaches mostly try to consider what the user

requires without considering the proﬁle of the user. In

this paper we present a discovery approach involving

the user proﬁle as a functional requirement of the ser-

vice to be matched as well. Answers to the questions

such as how a goal can evolve in order to make dis-

covery more accurate, and how the underlying storage

architecture supports discovery.

The structure of this paper is as follows: Section 2

provides the motivating example, which we use to

present our approach of service discovery. Section 3

outlines modeling of web services and goals and Sec-

tion 4 presents the how the discovery of web services

takes place and how goal evolves during the process.

This section also describe the peer-to-peer architec-

ture on which the discovery component is relying on.

Section 5 describes the tools used to implement the

whole architecture, a run through example for the case

study is elaborated in section 6. Section 7 presents

some of related work done in this direction. Finally,

Section 8 draws some conclusions and presents future

directions.

2 MOTIVATING EXAMPLE

In this section we will present a motivating scenario

for this research work. In this example different va-

riety of driving licences are issued depending on the

Nazir S., Sapkota B. and Vitvar T. (2008).

PERSONALIZED WEB SERVICE DISCOVERY.

In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 82-89

DOI: 10.5220/0001526400820089

 SciTePress

person’s physical ability and the type of vehicle he

drives. The input required from the user is a type of

vehicle, whether he holds international/EU-State driv-

ing licence, and if he has any disability. The personal

proﬁle input would allow the proper discovery of re-

quired service(s). If in this case user proﬁle isn’t con-

sidered then a large amount of services will be discov-

ered from which the selection would have to be done

by the user.

3 SERVICE AND GOAL

MODELLING

In this section we present language and technique

used to model the web services and user goals. For

describing web services and user goals WSML-Flight

variant of WSML (Toma, 2007) is used. WSML-

Flight allows writing simple logical expressions and

rules. Web services, goals and ontologies are mod-

eled using WSMO (Roman et al., 2007).

The reason for modeling and using personaliza-

tion is to provide different discovery support from

user to user, depending on his needs, status and

environment, rather than providing crude services.

Some work towards personalization for information

retrieval has been done in digital libraries (Amato and

Straccia, 1999), where users are modeled in terms of

ﬁve categories, which are personal data, preferences

and restrictions, time and location of delivery, high re-

sponsiveness of the service to user’s need and privacy

of the user data. Moving towards the web, (Garrigos

et al., 2005), Garrigos has presented reusable mod-

els for personalization. In their approach ﬁve models

are presented. Domain model to deﬁned the struc-

ture of domain data, Navigation model to deﬁne struc-

ture and behavior over the domain data. Presentation

model is used to handle the layout of the presenta-

tion. The personalization model holds rules for dif-

ferent personalization policies and User model allows

to store data needed for the personalization.

In this paper we model the user in terms of his

needs and proﬁle. QoS parameters i.e price, delivery

time etc are not handled in our model, but are con-

sidered to be a non-functional property of the service

which the user can negotiate later.

We have modeled the need and proﬁle of the user

as a functional property of the service and the goal.

Using WSMO for modeling service and user goal,

both the proﬁle and the need will be in the funca-

tional description. The proﬁle of the user is deﬁned

as a concept in a domain ontology. The reason for

having this concept in domain ontology is the vary-

ing requirement of proﬁle for the domains. User of

the service can be an individual or an organization. A

service may be a user of another service, but the pro-

ﬁle used will be of the external entity. Some proﬁle

inputs may be required for one domain, while those

might not be of interest in the other. The service is

also dependent on the location of the user, e.g the

driving licence service is provided by many countries,

but the location input provided by the user will allow

providing a more accurate discovery.

A web service W can be modeled in WSMO as

follows:

W = {F, B}

F = {Wρ,W ρ

}

B = {σ,σ

}

F represents functional description of the web ser-

vice and B represents the behavioral description of the

web service. A functional description of web service

contains precondition Wρ and postcondition Wρ

a web service. WSMO provides two more functional

descriptions attributes which are assumption and ef-

fect. But we are only using precondition and postcon-

dition.

The behavioral description of the service B con-

tains choreography σ and orchestration σ

. The chore-

ography of a web service describes the communica-

tion pattern and how the service can be used. Orches-

tration represents how the overall functionality can be

achieved by cooperation of different web service.

The precondition of web service contains rules

which allow to verify the user’s proﬁle. The proﬁle

concept created in the domain ontology is used in the

precondition to evaluate the state of the information

before service execution. The precondition can be de-

ﬁned as follows:

W ρ = α

Where

α = {ν1, ν2,ν3,··· ,νn}

Here α is the set of conditions. This set contains

individual conditions νi which needs to be fulﬁlled

before the execution of the service. A service can

have multiple preconditions; all of them are need to be

fulﬁlled by the user goal. Starting from very generic

precondition and moving towards speciﬁc precondi-

tion for the services. The generic precondition is held

by all the services that provide same output. Moving

from generic preconditions towards specialized pre-

conditions allows differentiating the proﬁle input re-

quired by the various service of the same type.

Service type is basically what a service provides.

Web services are said to belong to the same service

type if they have a common postcondition, but a vary-

ing precondition.

This will allow to repeatedly evolving a goal

which can accurately match with a web services or

PERSONALIZED WEB SERVICE DISCOVERY

a set of web services.

The postcondition of the web service description

deﬁnes the state of the information that is reached af-

ter the service is executed. The need of the user is

modeled in the postcondition of the service. There

can be web service which have common postcondi-

tion but the precondition they require to be veriﬁed

are different. Postcondition can be deﬁned as follows:

W ρ

= {β}

Here β represents the state of the information after

execution of the service.

Beneﬁt of using WSMO for modeling is that it

provides support for modeling the user’s request. The

user’s request is divided into two parts, as need of the

user and proﬁle of the user.

We follow a similar approach for modeling goal

as presented in (Stollberg et al., 2007). We use goal

template and goal instance. Goal template can be de-

ﬁned as follows:

= {Gρ,Gρ

}

A goal template G’ contains precondition Gρ and

postcondition Gρ

. Precondition of a goal template is

the generic precondition which all the services in one

service types deﬁne.

Following is how goal is deﬁned, Let G represent

the WSMO goal:

G = {Gρ

}

Here Gρ

is the postcondition of the goal. User’s

need is captured by the postcondition of the goal. We

are not using precondition in the functional part of

the goal. The reason of not using precondition is dif-

ﬁculty for the user to write logical expressions. The

user is not interested deﬁning the conditions on the

information but rather would like to provide his input

as simple data. Thus for supporting the user to deﬁne

his proﬁle, we use the user’s proﬁle concept described

in the domain ontology.

The user request then ﬁnally comprises of WSMO

goal containing postcondition and user data as user’s

proﬁle concept.

Goal templates are provided by service provider

and are created at design time. Transition from goal

template to goal takes will be explained in the goal

creation section presented later. But for now, goal

template is used as a starting point for creating a goal

instance.

4 ARCHITECTURE OVERVIEW

4.1 Goal Creation and Service

Discovery

This section elaborates the user’s perspective of the

overall architecture and how a user’s request is for-

malized which is later used for discovery. WSMO

provides support for modeling a goal, but it does not

provide the technique for creating such a goal. An

effort made towards this technique is presented in

(Stollberg and Norton, 2007) . In this paper Stollberg

has presented two forms of goals namely Goal tem-

plates and Goal Instance. He makes further classiﬁca-

tion in Goal templates, which are basic type and com-

posed type. Our work is some what along the same

lines. Here we currently only consider goals of basic

type. Goal will be modeled using the goal modeling

technique presented in the previous section. The goal

evolves for an accurate discovery of service. As de-

ﬁned in the previous section a goal template contains

precondition from the generic type which belong to

all the web service of the same type. A user requests

for a web service goal template using key words. The

matching process retrieves the goal templates which

match his request. User input is required for all the

precondition variables. The precondition is used to

create a user proﬁle which is sent with the goal in a

form of instance of the concept. Evolution of goal

takes place during the web service discovery process.

Each partial input is matched with a complete descrip-

tion of web service. Incomplete inputs become a part

of precondition and are requested in the next step. The

process can be represented as follows: Let initial par-

tial goal created from goal template be

GoalTemplate = G

{Gρ,Gρ

}

Gρ = α{ν1/t1}

Here ν’s are free variables. These variables are

replaced by the t input given by the user.

User Data {δ} ⊂ α{ν1/t1}

This precondition results into creation of user data

as the concept of a proﬁle in an ontology. This con-

cept satisﬁes the precondition of the goal template.

From a goal template a goal is created having only

the postcondition of the goal template. Input for dis-

covery is ρ

,O(δ) goal with postcondition and ontol-

ogy containing user data proﬁle concept. Through the

discovery process user proﬁle is reﬁned. The pre-

condition not matching is returned from the discov-

ery component. Just as the case with goal template

this precondition is used to take input from user to ex-

tend the user proﬁle. Next the precondition with this

extended data is sent again to discovery component.

WEBIST 2008 - International Conference on Web Information Systems and Technologies

This process continuous until all the precondition of

the service is satisﬁed. The following relation must

hold:

Gρ ⊂ W ρ

Since

Gρ ⊂ α{ν1/t1,...,νn/tn}

And

δ ⊂ α

Then

δ ⊂ W ρ

This relation explains that the user data should

hold for the precondition of the web service. The

postcondition of the goal must also hold for the web

service postcondition. Following is the relation for

postcondition.

Gρ

⊂ W ρ

The process of discovery ﬁrst matches postcondi-

tions. Web services which have matching postcon-

ditions are then taken for matching the preconditions

of the service. The output from the discovery com-

ponent is list of web service along with precondition

which need to be satisﬁed. If the web service has no

more preconditions to be matched then the second at-

tribute is null. From the list the user selects which web

service fulﬁls his need. On selecting a service with

un-fulﬁlled precondition, then the user has to provide

input.

Here the deviation from the traditional approach

is taken, where the following relation must hold:

Gρ ⊂ W ρ

Gρ

⊂ W ρ

Creating an instance of a concept from the input

provided by the user is simpler. This would save the

user from writing complex logical expressions and

for component to manipulate with these expressions,

without complete knowledge of the domain. This in-

stance data will allow creating simple queries from

the web service description to verify if the instance

holds.

4.2 Storage Requirement

In this section we present the storage architecture of

our discovery approach. The underlying storage ar-

chitecture is based on peer to peer architecture as pre-

sented in (Sapkota et al., 2007). Service providers

/ agency providing the web service description acts

as a peer and becomes a part of the semSet which

is of the same domain. There are few changes made

in the overlay network, to cope with the requirement

of the e-gov domain. The domains of the service

are clearly deﬁned. The peer can belong to multiple

domains. Another e-gov domain speciﬁc restriction

is the privacy of the member state i.e to keep data

Figure 1: Peer-to-Peer Architecture for discovery.

of one member state unaccessible to the other mem-

ber state. There is a communal gateway which pro-

vides mediation between ontology. Semantic query

(WSMO goal) requiring interaction with other mem-

ber state passes through the communal gateway. More

detail regarding the communal gateway can be found

in (Vitvar et al., 2006). The mediation approach

used can be found in (Mocan, 2005), which is a part

of WSMX architecture (Zaremba and Moran, 2005).

WSMX (web service execution environment) is run-

ning on all nodes. The discovery and storage compo-

nents are part of WSMX architecture. Each member

state has its own coSet. From the member state coSet

one of them becomes the member of inter-member

state level coSet. Inter-state querying takes place

through these peers. For details about the construc-

tion and membership algorithm of the network refer

to (Sapkota et al., 2007). In ﬁgure. the doted line

ring represent the semSet. Where there is only one

peer in the semSet, that peer is a member of the mem-

ber state coSet. The dashed black line represents the

member state coSet.

5 IMPLEMENTATION

The discovery component provides the functionality

of discovery of semantic web services which are most

suitable for the user. Figure 2, shows structure of each

peer. Discovery component performs matchmaking

between WSMO goal and proﬁle provided as an in-

stance of citizen concept with the semantic service

PERSONALIZED WEB SERVICE DISCOVERY

descriptions. Personalization component handles the

ordering of personalization rules in the preconditions

needed to be veriﬁed by the discovery component,

and returns the formatted output to the external com-

ponent. The Repository holds the semantic data and

communication engine handles the peer-to-peer com-

munication.

Figure 2: Peer Structure.

The implementation has been done in Java 1.5.

We have used the WSML2Reasoner framework

(Motik et al., 2005), and KOAN2 reasoner for eval-

uation of logical expressions deﬁned in the seman-

tic descriptions of the services. The underlying

communication between the peers is handled by

JXTA (Li, 2003). The semantic repository used is

ORDI (Ognyanov and Kiryakov, 2005).

6 RUN THROUGH

In this section we present a running example for the

case study. The requirement is that the user needs a

Irish driving licence. In this example we will consider

that the ontology being used is same for goal and web

service. Non-Functional property(nfp) includes the

domain and member state the entity belongs to. These

nfps are used for routing purpose between the peers.

Table 1 and Table 2 shows how the web service, goal

template and goal are described.

Figure 3, shows how the goal evolves and discov-

ery takes place. Initially the user is provided with a set

of keyword, which allows him to select what his need

is. Using these keywords goal templates are matched.

From the list of goal templates the user selects one.

Figure 3: Discovery Process.

The goal template as deﬁned in the modeling section

has precondition with free variables. Input is taken

from the user for these free variables. These input

result in the creation of instance of the person con-

cept deﬁned in the ontology. This instance along with

the goal having the postcondition only is sent to the

discovery component. The discovery component per-

forms matchmaking and as a result returns a web ser-

vice / set of web services, along with free variable

which require more user input. If the service has no

corresponding free variable it is considered to be dis-

covered. As the user selects a goal template from the

list of goal template, here again the user has to select

a web service from the list of web services, which

serves in the same way as goal template. This activity

continuous until no more free variable are available

along with the service, or the user is satisﬁed with the

discovered service.

In this case study, there are services which provide

different type of driving licences. Here just for sim-

plicity and to present our point we consider only one

type of licence. The concentration in this example is

to present how the goal evolves and how precondition

of the web service plays its part in the discovery pro-

cesses. The keyword based approach to select the goal

template is used. This goal template asks for user to

ﬁll the nationality, age and if the user has any physical

disability. This resulting in the creation of a goal with-

out precondition and instance data. Next this goal is

sent for discovery. The discovery component returns

”Web service A” that matches the goal and instance

WEBIST 2008 - International Conference on Web Information Systems and Technologies

Table 1: Web Services.

webService ”webServiceA.wsml”

W ρ = α{ν1, ν2}, W ρ

= {β} ν1 = ?req{Nationality =”IrishCitizenship”

∧ age ≥ 18 ∧ disable = ”false”} memberOf request.

ν2 = ?req{IrishId = ”true”} memberOf request.

β = ?res{VehicleType = ”car” } memberOf drivingLicence.

This web service provides a driving li-

cence of Ireland for a car. The generic

precondition for this service is that the

person is an Irish citizen and has no dis-

ability. For specialized precondition the

citizen holds an Irish id as well.

webService ”webServiceB.wsml”

W ρ = α{ν1,ν2}, W ρ

= {β}

ν1 = ?req{Nationality =”IrishCitizenship” ∧ age ≥ 18 ∧ disable =

”true”} memberOf request.

ν2 = ?req{IrishId = ”true”} memberOf request.

β = ?res{VehicleType = ”car” } memberOf drivingLicence.

This web service provides a driving li-

cence of Ireland for a car. The generic

precondition for this service is that the

person is an Irish citizen and has some

disability. For specialized precondition

the citizen holds an Irish id as well.

webService ”webServiceC.wsml”

W ρ = α{ν1,ν2}, W ρ

= {β}

ν1 = ?req{Nationality =”EU” ∧ age ≥ 18 ∧ disable = ”false”} mem-

berOf request.

ν2 = ?req{EUId = ”true”} memberOf request.

β = ?res{VehicleType = ”car” } memberOf drivingLicence.

This web service provides a driving li-

cence of Ireland for a car. The generic

precondition for this service is that the

person is an EU citizen and has no dis-

ability. For specialized precondition the

citizen holds and EU member state id as

well.

Table 2: Goal evolution.

goal ” goalTemplate.wsml”

Gρ = α{ν1}, Gρ

= {β}

ν1 = ?req { Nationality = ?nat ∧ age = ?age ∧ disable = ?dis} mem-

berOf request.

β = ?res{VehicleType = ”car” } memberOf drivingLicence.

This goal template provides a starting

point to discover a service which pro-

vides a deriving licence of Ireland for a

car. The precondition provide what in-

formation is required from the user. The

free variables are used to taking input.

The postcondition deﬁnes that the user

is looking for a web service which pro-

vides car driving licence. This goal tem-

plate is created at design time by the ser-

vice provider.

goal ” goalInit.wsml”

Gρ

={β}, O = {δ}, β = ?res{VehicleType = ”car”} memberOf driv-

ingLicence. δ = ?data{Nationality =”IrishCitizenship” ∧ disable =

”false” ∧ age = 20 } memberOf Citizen.

This goal is created after the user’s input

for the free variables.

The instance created has all the data

which is needed for discovery

goal ” goalFinal.wsml”

Gρ

={β}, O = {δ}, β = ?res{VehicleType = ”car”} memberOf driv-

ingLicence. δ = ?data{Nationality =”IrishCitizenship” ∧ age = 20 ∧

disable = ”false” ∧ IrishId = ”true”} memberOf Citizen.

This is the second level of the goal,

where after the ﬁrst iteration web ser-

vice discovered, i.e. ”web service A”

requires further input for precondition.

To check if the person holds an Irish Id

card.

data, which is if the user has the Irish Id. This pre-

condition like the goal template is used to take input

from the user and to update the instance data. This ﬁ-

nal goal and instance data is sent for discovery which

ﬁnally only returns ”Web service A” with null for fur-

ther input requirement. The routing of the goal takes

place using the nfps. Once the peer receives a request

for discovery, it uses the nfps to check to which peers

to route the query to. Peer receiving the request has

in its routing table peers who belong to that domain

and member state, it forwards the goal to those peers.

In case the peer holds no information for that domain

or member state, it forwards the request to the coordi-

nator peer. This forwarding of request continues until

the inter-member state coordinator is reached. The

coordinator forwards the request to the member state

to which the query belongs to. The result is returned

along the same path taken to reach the node.

7 RELATED WORK

There has been a lot of work done for web service dis-

covery. Most of the work concentrates on the simple

web service discovery without considering the con-

PERSONALIZED WEB SERVICE DISCOVERY

text or the personalized view of the user.

The two phase discovery presented in (Stollberg

et al., 2007), which we extend to create goal for dis-

covery, provides a way to model the goal for discov-

ery and discover web service, but does not take into

consideration about the possible use of precondition

for providing accurate web services.

Wolf-Tilo presents a personalized selection of web

services approach in (Balke and Wagner, 2003). The

methodology presented there handles the task of per-

sonalization at selection time. The discovery ap-

proach used in keyword based. Ther discovery pro-

posal lacks expressiveness of the query and the ser-

vice. They use simple SQL query to select the ser-

vices that meet the user’s requirement.

(Maamar et al., 2006), gives a very compre-

hensive view for modeling the personalization aspect

for web service discovery. They present a discov-

ery methodology which requires interaction with the

web services and the resource where the web service

has to be executed. They consider three aspects when

considering about the personalization, which are User

context, web service context and resource context.

User context considers user’s preferences, including

where and when the service is needed to be executed.

The web service context considers the constraints on

the web service to execute in a speciﬁc environment

and lastly the resource context is the time and load

constraint on the resource. Their approach is quite

similar to our in terms user’s view, but not much de-

tail is presented regarding the modeling language of

the service and query from the user.

8 CONCLUSION AND FUTURE

WORK

In this paper an approach for discovering web services

has been presented. This papers considers discovery

from users perspective. Creation of reﬁned goal from

a generic one is provided to make an accurate service

discovery. Modelling of web services and goal has

been provided using WSMO. WSMO allows to model

a proﬁle of the user as ontology concept. We have pre-

sented how the use of precondition of the WSMO web

service description can be done, which earlier has not

been used. A distributed architecture has been pre-

sented, which handles distributed querying. This has

resulted in efﬁcient discovery of services as the load

on the discovery component has been distributed. The

inferencing time required for discovery is reduced.

This architecture has been implemented in the EU

funded SemanticGov, and has result in very accurate

service discovery during its testing. More work needs

to be done for handling composed goals. Creating

sub-goals and handle discovery and aggregating re-

sults. This has been left for the future work.

ACKNOWLEDGEMENTS

This material is based upon works supported by the

EU funding under the projects TripCom (FP6-IST-4-

027324-STP), SemanticGov (FP6-IST-4-027517) and

by the Science Foundation Ireland under Grant No.

SFI/02/CE1/I131.

REFERENCES

Amato, G. and Straccia, U. (1999). User proﬁle model-

ing and applications to digital libraries. In Proc. of

3rd European Conference on Research and Advanced

Technology for Digital Libraries, pages 670–683.

Balke, W. and Wagner, M. (2003). Towards personalized

selection of web services. In Proc. of 12th World Wide

Web Conference.

Clement, L., Hately, A., Riegen, C., and Rogers,

T. (2004). Uddi version (3.0.2), uddi spec

technical committee draft. Technical report.

http://www.w3.org/TR/2004/REC-owl-features-

20040210/.

Garrigos, I., Gomez, J., Barna, P., and Houben, G. (2005).

A reusable personalization model in web application

design. In Proc. of Workshop on Web Information Sys-

tems Modelling, pages 40–49.

Hai, L., Qing, L., and Naijie, G. (2005). Quantifying con-

texts for user-centered web service discovery. In Proc.

of 10th IEEE International Enterprise Distributed Ob-

ject Computing Conference, pages 399–404.

Keller, U., Lara, R., Polleres, A., Toma, I., Kifer, M., and

Fensel, D. (2004). D5.1 v0.1 WSMO Web Service Dis-

covery WSML Working Draft.

Li, S. (2003). Jxta 2: A high-performance, massively scal-

able p2p network. Technical report.

Maamar, Z., Mostefaoui, S., and Mahmoud, Q. (2006).

Context for personalized web services. In Proc. of

38th Hawaii International Conference on System Sci-

ences.

Makripoulias, Y., Makris, C., Panagis, Y., Sakkopoulos, E.,

Adamopouloua, P., and Tsakalidis, A. (2006). Web

service discovery based on quality of service. In Proc.

of 4th ACS/IEEE International Conference on Com-

puter Systems and Applications, pages 196–199.

Mocan, A., editor (2005). WSMX Data Mediation. WSMO

Working Group.

Motik, B., Nagypal, G., and Grimm, S. (2005). Wsml rea-

soner. Technical report.

Ognyanov, D. and Kiryakov, A. (2005). Ontology represen-

tation and data integration (ordi) framework. Techni-

cal report.

WEBIST 2008 - International Conference on Web Information Systems and Technologies

Paolucci, M., Kawamura, T., Payne, T., and Sycara, K.

(2002). Semantic matching of web service capabil-

ities. In Proc. of 1st Int. Semantic Web Conference,

pages 333–347.

Roman, D., Lausen, H., and Keller, U., editors (2007). The

Web Service Modeling Ontology (WSMO). WSMO

Working Group.

Sapkota, B., Nazir, S., Vitvar, T., Toma, I., Vasiliu, L., and

Hauswirth, M. (2007). Semantic overlay for scal-

able service discovery. In Proc. of 3rd International

Conference on Collaborative Computing: Network-

ing, Applications and Worksharing.

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

owl-s to uddi, implementation and throughput. In

Proc. of 1st International Workshop on Semantic Web

Services and Web Process Composition (SWSWPC).

Stollberg, M., Keller, U., Lausen, H., and Heymans, S.

(2007). Two-phase web service discovery based on

rich functional descriptions. In Proc. of 4th European

Semantic Web Conference, pages 99–113.

Stollberg, M. and Norton, B. (2007). A reﬁned goal model

for semantic web services. In Proc. of 2nd Interna-

tional Conference on Internet and Web Applications

and Services, pages 17–22.

Toma, I., editor (2007). The Web Service Modeling Lan-

guage (WSML). WSMO Working Group.

Vitvar, T., Kerrigan, M., Overeem, A., Peristeras, V., and

Tarabanis, K. (2006). Infrastructure for the seman-

tic pan-european e-government services. In Proc. of

AAAI Spring Symposium on The Semantic Web meets

eGovernment (SWEG).

Vu, L., Hauswirth, M., and Aberer, K. (2005). Towards

p2p-based semantic web service discovery with qos

support. In Proc. of 3rd Business Process Manage-

ment Workshops, pages 18–31.

Zaremba, M. and Moran, M., editors (2005). WSMX Archi-

tecture. WSMO Working Group.

PERSONALIZED WEB SERVICE DISCOVERY