IMPROVED SERVICE RANKING AND SCORING: SEMANTIC

ADVANCED MATCHMAKER (SAM)

Erdem S. Ilhan, Gokay B. Akkus

Department of Computer Engineering, Bogazici University, Istanbul, Turkey

Ayse B. Bener

Department of Computer Engineering, Bogazici University, Istanbul, Turkey

Keywords: Matchmaking, semantic similarity, scoring, ranking, OWL-S, bi-partite graph, scoring, semantic distance.

Abstract: In recent years Semantic Web has drawn a lot of attention in order to solve the problem of automatic

discovery and processing of web services. Although there are different efforts and frameworks for semantic

annotation and discovery of web services, they mostly classify the discovered web services as set-based.

Improvement in matching process could be gained by the use of ontological information in a useful form.

The goal of this research is to propose a more accurate discovery method using the ontological distance

information defined and ranked by users. In this paper, we focus on one of the most challenging tasks in

service discovery: matchmaking process. We use an efficient matchmaking algorithm based on bi-partite

graphs. Our proposed algorithm uses attribute ranking through weight assignment. Our experiment results

show that bi-partite matchmaking has advantages over other approaches in the literature for parameter

pairing problem. We present value added approaches in matchmaking such as property-level matching,

semantic distance information and WordNet scoring. The value added approaches provide better scoring

scheme and allows similarity to be captured resulting in ranking of services according to their relatedness.

1 INTRODUCTION

In recent years, web services became the dominant

technology in providing the interoperability among

different systems throughout the web. The problem

of finding the right and most suitable web services

for user needs emerges when open e-commerce

systems are widely used and user requirements

dynamically change over time.

Although there are currently proposed

technologies for discovery of web services, such as

UDDI (http://www.uddi.org,, 2006.), they do not

satisfy the full discovery requirements. This

discovery process is based on syntactical matching

and keyword search that does not allow the

automatic processing of web services. To solve the

problem of automatic discovery and processing of

web services, the Semantic Web (

http://www.w3.org/2001/sw/, 2006.) as a new vision

is proposed. The Semantic Web is an effort by the

W3C consortium (http://www.w3.org/, 2006.), and

one of its main purposes is to facilitate the discovery

of web resources.

There are different efforts and frameworks for

semantic annotation and discovery of web services

(Motta, E., J. Domingue, L. Cabral and M. Gaspari,

2003, Fensel, D. and C. Bussler, 2002). For web

service discovery they also propose some techniques

and algorithms. However, they mostly classify the

discovered web services in set-based approaches.

They do not focus on rating the web services using

semantic distance information (Klein, M., B. Konig-

Ries and M. Muussig, 2005).

The evolution of web services, from

conventional services to semantic services, caused

service descriptions contain extra information about

functional or non-functional properties of web

services. The semantic information included in the

service descriptions enables the development of

advanced matchmaking schemes, capable of

assigning degrees of match to the discovered

services. Semantic discovery of web services means

semantic reasoning over a knowledge base, where a

S. Ilhan E., B. Akkus G. and B. Bener A. (2007).

IMPROVED SERVICE RANKING AND SCORING: SEMANTIC ADVANCED MATCHMAKER (SAM).

In Proceedings of the Second International Conference on Evaluation of Novel Approaches to Software Engineering , pages 95-102

DOI: 10.5220/0002586700950102

 SciTePress

goal describes the required web service capability as

input. Semantic discovery adds accuracy to the

search results in comparison to traditional Web

service discovery techniques, which are based on

syntactical searches over keywords contained in the

web service descriptions (U. Keller, Lara R.,

Polleres A, 2004).

Improvement in matching process could be

gained by the use of ontological information in a

useful form. With the use of this information, it can

be possible to rate the services found in discovery

process. As in real life, users/ agents should be able

to define how they see the relation of ontological

concepts from their own perspective. Similarity

measures have been widely used in information

systems (Voorhees, E, 1998, Ginsberg, A., 1993,

Lee, J., M. Kim and Y. Lee, 1993), cognitive

science, software engineering and AI (Agirre, E. and

G. Rigau, 1996, Hovy, E., 1998, Wang, Y. and E.

Stroulia, 2003). So integration of knowledge from

these techniques can improve the matching process.

By using semantic distance definition

information, we aim to get a rated and ordered set of

web services as the general result of the discovery

process. We believe that this would be better than

set-based classification of discovered services. In

this paper, we propose a new scheme of

matchmaking that aims to improve retrieval

effectiveness of semantic matchmaking process. Our

main argument is that conventional evaluation

schemes do not fully capture the added value of

service semantics and they do not consider the

assigned degrees of match, which are supported by

the majority of discovery engines. The existing

approach to service matchmaking contains

subsumption values regarding the concept that the

service supports. In our proposed approach, we add

semantic relatedness values onto existing

subsumption-based procedures. We introduce value

added approaches to matchmaking process such as

property-level matching, semantic distance

information and WordNet scoring. Property-level

matching provides capturing similarity between

concepts that do not have a subsumption relation. So

that, services that would not be classified, are ranked

with our matchmaking agent. Similarity distance

provides user’s profile to be represented in the

ontology. Similarity distance weights can be

assigned on the links between concepts to specify

concepts relatedness to each other in an explicit way.

Also by making use of WordNet, we introduce a

second source of semantic repository to be utilized

in matchmaking. Our test results in section 5

indicate that these value added approaches increases

the captured semantic relations between parameters

of services and provide a better ranking of services

resulting in better user experience in matchmaking.

2 RELATED WORK

Semantic Web services aim to realize the vision of

the Semantic Web, i.e. turning the Internet from an

information repository for human consumption into

a worldwide system for distributed Web computing

(http://www.w3.org/2001/sw/, 2006.). The system is

a machine-understandable media where all the data

is combined with semantic metadata. The domain

level formalizations of concepts form up the main

element within this system, which is called ontology

(http://www.w3.org/Submission/OWL-S, 2004).

Ontology represents concepts and relations between

the concepts; these can be hierarchical relations,

whole-part relations, or any other meaningful type of

linkage between the concepts(H. El-Ghalayini, M.

Odeh, R. McClatchey, and T. Solomonides,, 2005).

The semantic matchmaking process is based on

ontology formalizations over domains. In the

upcoming section we present some of the selective

research on the matchmaking process considering

the concepts that we build our research on.

Matchmaking of Web services considers the

relationship between two services. The first one is

called the advertisement and the other is called the

request (Klusch, M., Fries, B., Khalid, M., and

Sycara, K.. 2005). Advertisement denotes the

services description of the existing services while

the request indicates the picture of service

requirements (Wang, Y. and E. Stroulia, 2003).

In (Wang, H., Zengzhi L., Fan L., 2006), the

problem of capability matchmaking is analyzed with

regarding to Web services, especially the

Preconditions and Effects (PE) matchmaking. In the

paper, the authors present a service similarity

function that determines similar parameter classes

by using a matching process over synonym sets,

semantic neighbourhood, and distinguishing

features. Parameter pairing is the process that is used

for matching service descriptions. In the work,

maximum weight bi-partite graph matching method

is utilized for parameter finding; the weights of bi-

partite graph’s edges are evaluated with matching

degree between function parameters calculated by

the similarity function mentioned above.

Although good results are obtained with the

usage of this method, it should still be improved in

two terms: One is that, it needs to be extended on

pre-condition and affect because the matching is

ENASE 2007 - International Conference on Evaluation on Novel Approaches to Software Engineering

performed only on parameters of input and output,

and the functional signature is not sufficient to

identify what it does. The other is that this

framework should be combined with particular

directory service like UDDI in order to improve the

discovery efficiency.

In (Paolucci M.; Kawamura,T.; Payne,T. and

Sycara,K. , 2002) the authors present an algorithm

that deals with the localization of Web services. The

research does not address the interoperability

problem. The system introduced uses the service

profile ontology from the DAML-S specification

language but only considers the matching of input

and output concepts defined by the same ontology.

Traditional approaches to modelling semantic

similarity between Web Services compute subsume

relationship for function parameters in service

profiles within a single ontology. In (Ruiqiang Guo,

Dehua Chen, Jiajin Le, 2005) a graph theoretic

framework based on bi-partite graph matching for

finding the best correspondences among function

parameters belonging to advertisement and request

is introduced. On computing semantic similarity

between a pair of function parameters, a similarity

function is introduced, determining similar entity,

which relaxes the requirement of a single ontology

and accounts for the different ontology

specifications. The function presented for semantic

similarity across different ontologies provides an

approach to detect similar parameters. The method is

based on a matching process over weighted sum of

synonym sets, semantic neighbourhood, and

distinguishing features. The method mainly lacks

use of functional similarities and lexical evaluation

of semantic mappings.

In (COMPSAC, 2006), a semantic ranking MSC

is designed to rank the results of advertisements

matchmaking. MSC stands for the initials of three

factors’ second words: Semantic Matching Degree

(to capture the semantic aspects of attributes),

Semantic Support (to describe the interestingness or

potential usefulness of an attribute) and Relational

Confidence (to capture the association relationships

among attributes). Three categories of attributes are

defined in advertisements matchmaking:

Generalizable Nominal Attribute (GNA) whose

values can form a concept hierarchy; Numeric

Attribute (NUA), called quantitative attribute, whose

values are numeral; Nominal Attribute (NOA)

whose values are neither numeral nor can form a

concept hierarchy. Three new factors are designed to

capture the semantic characteristics and relationships

of the attributes: Semantic Matching Degree,

Semantic Support and Relational Confidence.

3 PROBLEM STATEMENT

The first step in service composition is identifying

the domain of interest by means of taxonomy of

subject categories. The discovery and selection of

services those are suitable for a given request is

obtained in two phases. Firstly, matchmaking

approach is based on the ontological framework. It is

applied on the set of available services in order to

find services that match needs of the requestor from

a functional point of view. Secondly, services are

ranked and further refined. This is done by taking

into account context information of the requestor.

Then preconditions or post-conditions can be

defined as mandatory or optional.

The problem that we are concerned is the first

step of this scenario: given a request r, finding right

web services for r. The main goal of this research is

to gain better precision and recall values on

matchmaking by considering user requests in web

services discovery.

The previous work on semantic matchmaking

focused on taking advantage of a single

implementation based on some information retrieval

theory. The experimental research so far has shown

simple subsumption based matchmaking is not

sufficient to capture semantic similarity.

In this research, we aim to provide an efficient

and accurate matchmaking algorithm using scoring

and ranking based on similarity distance

information, extended subsumption and property

level similarity assessment in a general semantic

web service discovery framework.

4 PROPOSED SOLUTION

In this paper we propose a hybrid approach on

semantic matchmaking. Our proposed solution uses

decision modules that can be plugged in and out. We

have implemented some of these modules to add

semantic relatedness values onto existing

subsumption-based procedures. Our proposed

matchmaker agent mainly provides ranking and

scoring based on concept similarity. The

components of the proposed system are shown in

Figure 1. Request service definition and the

corresponding relevant services set, which are

discovered through conventional discovery

mechanisms, are presented as input to the system.

The ontology and services we use are retrieved from

“OWL-S Service Retrieval Test Collection version

2.1”. The services in the collection are mostly

extracted from public UDDI registries, providing

IMPROVED SERVICE RANKING AND SCORING: SEMANTIC ADVANCED MATCHMAKER (SAM)

582 web services described in OWL-S from seven

different domains. The OWL-S Test collection

version 2.1 contains 29 queries, each of which

associated with a set of 10 to 15 services (Mahboob

Alam Khalid, Benedikt Fries, Patrick Kapahnke,

2006). We extended some ontologies in this test

collection for our own purposes in order to better

demonstrate the features of our proposed

matchmaking agent. We believe that a formal test

collection of OWL-S services is crucial for the

evaluation of matchmaking agents.

Figure 1: Matchmaking agent components.

The main software components of our proposed

matchmaking agent are shown in Figure 2. The top

layer represents our matchmaker SAM (Semantic

Advanced Matchmaker). OWL-S API models the

service, profile, process and grounding ontologies of

OWL-S in an easy to use manner. It is a widely used

API in semantic applications. OWL-S API also

presents interfaces for reasoning operations and

utilizes Jena constructs at the back-end. At the

bottom of the hierarchy we have Pellet reasoner for

OWL reasoning operations.

Figure 2: Software components of matchmaking agent.

We believe that a discrete scale (exact, plug-in,

subsume, and fail) of service classification is not

sufficient for a matchmaking process. On the other

hand, semantic ranking of services can capture a set

of services that are lost in a discrete scale match.

Semantic similarity assessment is a crucial step for

the ranking process. In our proposed system, we

present value-added similarity assessment

approaches between service and request parameter

pairs.

4.1 Matching Algorithm

Previous research has shown that bi-partite graph

matching algorithm is a good fit for finding

matching parameters in a service and request pair

(Herbert Alexander Baier Saip, Claudio Leonardo

Lucchesi, 1993). Bi-partite graph matching provides

us a solution for parameter pairing problem. We

consider the inputs and outputs as separate cases and

partition the service parameters and request

parameters to form the bi-partite graph. The

similarity assessment process of our matchmaker

assigns weights for each parameter pair on this bi-

partite graph. A maximum weight match on the final

graph leaves us with the optimum matching

parameter pairs and with a score that is sum of the

weights between matched parameter pairs. We

repeat this process for each service and request pair

and finally rank the services according to their score

from bi-partite graph matching algorithm.

As we stated before the process that

differentiates the services is the similarity

assessment process. We consider OWL-S profiles of

service definitions and assign similarity scores for

input and output parameter pairs. We present the

following value-added features for similarity

assessment: Subsumption based similarity, WordNet

based similarity, similarity distance information and

WordNet similarity assessment.

4.1.1 Subsumption based Similarity

Assessment

We make use of OWL-DL constructs subClassOf,

disjointWith, complementOf, unionOf and

intersectionOf to assess concept similarity based on

subsumption. If two concepts are explicitly stated to

be complement or disjoint, a zero score is directly

assigned. Otherwise, we check for subclass relation

and also assess according to property level

assessment procedure described below.

We wanted to capture similarity values in bi-

partite graph since it is important to decompose

concepts that include the characteristic of “a union

of”. Following this approach, we always pair and

assess score for atomic concepts in matchmaking

process.

ENASE 2007 - International Conference on Evaluation on Novel Approaches to Software Engineering

4.1.2 Property-level Similarity Assessment

We have assumed that in matchmaking it is also

important to have properties and their associated

range in measuring the degree of match. Such as, if

two concepts have similar properties (properties

having subclass relation) and their range classes are

similar, then this improves their level of similarity.

Using property level similarity assessment ranks a

service that would normally be eliminated by a

conventional matchmaker. For example, a user

request may favour a particular author for a novel. A

service, which returns articles that are written by that

particular author, will have a high score even though

the concept of “an article” does not compare to the

concept of “a novel”. Therefore our proposed system

returns positive results for concepts that have similar

properties as well as the similar concepts.

4.1.3 Similarity Distance based Assessment

To represent similarity distance information we

applied N-ary relation pattern in OWL, which is

used to represent additional attributes on a property.

The additional attribute in our case is the similarity

distance value. Figure 3 shows how this pattern is

organized:

Figure 3: N-ary relation pattern in OWL, representing

similarity distance information.

SimilarityRelation concept is introduced as a

class with this pattern and the similarity distance

value is represented as the range of

hasSimilarityDegree property of this concept. The

similar classes are represented as Concept_1 and

Concept_2 in Figure 3.

We follow the standards approach by

representing similarity distance information in

OWL, which can be imported and used in other

ontologies (Şenvar, M. and Bener, A., 2006).

Similarity distance information is useful in reflecting

user’s profile on the ontology. The importance and

relatedness of concepts for the user are represented

as weights on the ontology. In addition, if similarity

distance annotation is not found between two

concepts, then a default distance value is assigned

according to the following formula:

x,y

= 1/|subClassOf(x)

direct

| (1)

In the above formula Sd

x,y

represents similarity

distance between concepts x and y and

|subClassOf(x)

direct

| represents the number of

elements in set of direct subclasses of concept x.

4.1.4 WordNet based Similarity Assessment

WordNet organizes words into synonym sets, which

are also linked to each other representing a semantic

relation. In our matchmaker we take WordNet as a

secondary source of information with the ontology

repository. We aimed at reasoning with these highly

structured information sources in order to get more

reliable result sets.

We make use of wordnet::similarity open source

project to assess similarity score among words. The

path length criterion is used for score assignment.

The parameter types of services are presented as

input to wordnet::similarity module.

5 EVALUATION AND RESULTS

In order to evaluate the performance of our proposed

matchmaking agent we extended the book ontology

in OWL-S Service Retrieval Test Collection (OWL-

S TC) and also modified related request and service

definitions accordingly (Mahboob Alam Khalid,

Benedikt Fries, Patrick Kapahnke,, 2006). As shown

in Figure 4, we added subclasses of Magazine,

namely Foreign-Magazine and Local-Magazine

classes.

Figure 4: Printed Material ontology section.

As shown in Figure 5, we created subclasses of

Publisher: Ordinary-Publisher, Alternative-

Publisher and Premium-Publisher. We also created

Local-Author and Foreign-Author classes, which are

subclasses of class Author.

The matchmaking agent is developed in Java and

it makes use of open source semantic web libraries

like OWL-S API and Jena. We also used Pellet as

the reasoning engine for OWL operations.

IMPROVED SERVICE RANKING AND SCORING: SEMANTIC ADVANCED MATCHMAKER (SAM)

Figure 5: Person ontology section.

To represent subsumption reasoning, similarity

distance based assessment and property-level

similarity assessment capabilities we define the

following request and services as described in Table

Table 1: Test request and service set.

Inputs Outputs

Request Ordinary-Publisher,

Novel, Paper-Back

Local-Author,

Genre

Service 1 Publisher,

ScienceFictionBook

Author, Price

Service 2 Book, Alternative-

Publisher, Book-Type

Publisher,

Price, Date

Service 3 FantasyNovel, Author

Price, Comic

Service 4 Newspaper, Book-

Type, Person

Review,

Fantasy

Service 5 Publication, Book-

Type, Reader

Time,

Publisher

For the above test collection the property level

similarity assessment plays an important role. Even

though Novel concept has no subclass relation with

ScienceFictionBook concept, both concepts have

properties like writtenBy and publishedBy. Thus, our

matchmaker applies a subsumption reasoning on

ranges for these properties, which are Author with

its subclasses, and Publisher with its subclasses.

Finally, an additional score is provided for these

services. A conventional matchmaker would have

ignored these services as a “fail”.

The property level matching score is determined

by the following formula:

x,y

= w

*Subsumption

x,y

(2)

In the above equation, Sp

x,y

represents property

level match score between range concepts x and y.

We use the semantic score obtained through

subsumption, property level matching and semantic

distance. WordNet

x,y

represents the WordNet score

for concept names. The coefficients for subsumption

and WordNet are fixed at 0.9 and 0.1 after making

several experimental runs. We plan to apply a neural

network training approach to determine values for

coefficients utilizing a large training data in future.

The following equation represents how the

subsumption score and WordNet score is considered

as the final similarity score.

x,y

= w

sub

*Subsumption

x,y

word

*WordNet

x,y

(3)

x,y,

in the above equation,

represents final similarity

score between concepts x and y. Subsumption

x,y

represents semantic score obtained through

subsumption, property level matching and semantic

distance. WordNet

x,y

represents the WordNet score

for concept names. The coefficients for subsumption

and WordNet are fixed at 0.9 and 0.1 after making

experimental runs. We plan to apply a neural

network training approach to determine values for

coefficients utilizing a large training data in future.

To consider how semantic distance information

affects our ranking we introduced the following

weights into the book ontology as described in

following list:

Publisher → Ordinary-Publisher: 0.2

Publisher → Alternative-Publisher: 0.5

Publisher → Premium-Publisher: 0.3

Author → Local-Author: 0.3

Author → Foreign-Author: 0.7

Magazine → Foreign-Magazine: 0.7

Magazine → Local-Magazine: 0.3

Book → Short-Story: 0.2

Book → Science-Fiction-Book: 0.4

Book → Novel: 0.3

Book → Encyclopedia: 0.1

Novel → Science-Fiction-Novel: 0.6

Novel → Fantasy-Novel: 0.2

Novel → Romantic-Novel: 0.2

Book-Type → Hard-Cover: 0.7

Book-Type → Paper-Back: 0.3

The ranking with semantic distance information is

listed in Table 2 as follows:

Table 2: Service I/O similarity scores.

Input

Score

Output

Score

Overall

Score

Service1 0.916 0.143 0.452

Service 2 0.345 0.096 0.195

Service 3 0.896 0.0003 0.444

Service 4 0.148 0.0003 0.059

Service 5 0.187 0.096 0.075

Considering the inputs Service 1 got the greatest

score as it has a subsume relation with the request

parameter Ordinary-Publisher and property-level

matching with Novel parameter. The third request

parameter is not considered, as the service only

ENASE 2007 - International Conference on Evaluation on Novel Approaches to Software Engineering

100

needs two. For the outputs we have Service 1 ranked

higher than others. Indeed, none of the services

satisfy all output requirements of the request. But

considering similarity distance information the

ranking is determined as above. Overall score favors

the output score by assigning a higher weight to that,

as the outputs of a service is more important to the

requestor. As a result, Service 1 is the most related

service for the specified request.

The similarity distance formulation is defined as

follows:

x,y

= Sd

x,t

* Sd

t,k

*…*Sd

m,y

(4)

In the above equation, Sd

a,b

∈ [0,1] for any a and

b pair. Sd

x,y

represents similarity distance between

concepts x and y. The product of similarity distance

values on the path from x and y gives the value for

x,y

. If the concepts are not subclasses of eachother

then we take the path including their first common

parent in the hierarchy.

The final subsumption similarity score

considering the similarity distance is shown below:

x,y

= w

direction

* Sd

x,y

(5)

direction

in the above equation,

varies according to

the subsumption property. Considering input

parameters the service parameter to subsume the

request parameter is favored and in the case of

outputs the reverse is true. So, we set w

direction

coefficient to either 0.6 or 0.4 according to this

approach. The values are determined after running

experimental tests.

6 CONCLUSION

We proposed a novel advanced matchmaker, which

introduces new value-added approaches like

semantic distance based similarity assessment,

property level assessment and WordNet similarity

scoring. Instead of classifying candidate web

services in a discrete scale, our matchmaking agent

applies a scoring scheme to rank candidate web

services according to their relevancy to the request.

The ranking property enables to include some of

the relevant web services in the final result set

whereas they would have been discarded in a

discrete scale classification. Additionally, our

proposed matchmaking agent improves

subsumption-based matchmaking by utilizing OWL

constructs efficiently and by considering down to a

level of concept properties in the process.

We also introduced semantic distance annotation

in ontology to represent relevancy of concepts to the

user in a numerical way. Semantic distance

annotations improve the relevancy of returned web

service set as they actually represent user’s view of

ontology. WordNet similarity measurement is also

presented as a value-added feature, which acts as a

secondary source of information, strengthening the

power of reasoning.

Our experiment results show that property level

matching can be a good method to capture

similarities between concepts that do not have a

subsumption relationship. An improvement at this

point can be to consider similarity between

properties in addition to similarity of property range

objects. Besides, similarity distance information

provided us with a method to further differentiate

the importance of concepts from the point of view of

the user. The test results show how similarity

distance plays an important role in service ranking.

We are still working on other ontologies to

further test our matchmaker agent and plan to

consider preconditions and effects of services in

matchmaking process. This will require the use of

SWRL (Semantic Web Rule Language) to represent

preconditions and effects as rules in the system.

Another improvement will be to add context

aware decision-making capabilities, enabling our

matchmaking agent to reason based on user profiles,

preferences, past actions etc. The system that we

have presented can be considered as a basis for the

development of context-aware agent.

REFERENCES

Wang, H., Zengzhi L., Fan L., 2006. An Unabridged

Method Concerning Capability Matchmaking of Web

Services, In Proceedings of the 2006 IEEE/WIC/ACM

International Conference on Web Intelligence.

Klusch, M., Fries, B., Khalid, M., and Sycara, K., 2005.

OWLS-MX: Hybrid Semantic Web Service Retrieval.

In 1st Intl. AAAI Fall Symposium on Agents and the

Semantic Web, AAAI Press, Arlington VA.

U. Keller, Lara R., Polleres A., WSMO Web Service

Discovery, www.wsmo.org/2004/d5/d5.1/v0.1/20041112/.

H. El-Ghalayini, M. Odeh, R. McClatchey, and T.

Solomonides, 2005. Reverse Engineering Ontology to

Conceptual Data Models, In Proceeding (454)

Databases and Applications.

Universal Discovery Description and Integration

Protocol, http://www.uddi. org, 2006. Semantic Web,

W3C, http://www.w3.org/2001/sw/, 2006.

W3C, World Wide Web Consortium, http://www.w3.org/,

2006.

Mahboob Alam Khalid, Benedikt Fries, Patrick Kapahnke,

2006. OWL-S Service Retrieval Test Collection

Version 2.1, Deutsches Forschungszentrum für

Künstliche Intelligenz GmbH Saarbrücken, Germany.

IMPROVED SERVICE RANKING AND SCORING: SEMANTIC ADVANCED MATCHMAKER (SAM)

101

Herbert Alexander Baier Saip, Claudio Leonardo

Lucchesi, Matching Algorithms for Bi-partite Graphs,

Relatorio Tecnico DCC-03/93.

Motta, E., J. Domingue, L. Cabral and M. Gaspari, 2003

"IRS-II: A Framework and Infrastructure for Semantic

Web Services", Proceedings of the 2nd International

Semantic Web Conference (ISWC2003), Vol. 2870 of

LNAI., Springer, pp.306-318, Florida, USA. OWL-S

Submission,

http://www.w3.org/Submission/OWL-S, 2004.

Fensel, D. and C. Bussler, 2002. "The Web Service

Modeling Framework: WSMF", Electronic

Commerce: Research and Applications, Vol. 1, No. 2,

pp. 113-137.

Klein, M., B. Konig-Ries and M. Muussig, 2005. "What is

needed for semantic service descriptions? A proposal

for suitable language constructs", Proceedings of

Inernational Journal of Web and Grid Services, Vol.

1, No. 3/4, pp. 328-364.

Voorhees, E., 1998 "Using WordNet for Text Retrieval”,

C. Fellbaum(Editor),"WordNet: An Electronic Lexical

Database", pp. 285-303, The MIT Press, Cambridge.

Ginsberg, A., 1993 "A Unified Approach to Automatic

Indexing and Information Retrieval", IEEE Exper,

Vol. 8, No. 5, pp 46-56.

Lee, J., M. Kim and Y. Lee, 1993. "Information Retrieval

Based on Conceptual Distance in IS-A Hierarchies",

Journal of Documentation, Vol. 49, No. 2, pp. 188-207.

Agirre, E. and G. Rigau, 1996 “Word Sense

Disambiguation Using Conceptual Density”,

Proceedings of the 16th Conference on

ComputationalLlinguistics, Vol.1, pp. 16-22.

Hovy, E., 1998 “Combining and Standardizing Large-

scale, Practical Ontologies for Machine Translation

and Other Uses", Proceedings of the 1st International

Conference on Language Resources and Evaluation

(LREC), Granada, Spain.

Wang, Y. and E. Stroulia, 2003 “Semantic Structure

Matching for Assessing Web-Service Similarity”,

Proceedings of the 1st International Conference on

Service Oriented Computing, Trento, Italy.

Ruiqiang Guo, Dehua Chen, Jiajin Le, 2005. Matching

Semantic Web Services accross Heterogeneous

Ontologies, Proceedings of the 2005 The Fifth

International Conference on Computer and

Information Technology (CIT’05).

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

2002. Semantic matching of web services capabilities.

In Horrocks, I. And Hendler, J.eds.Proc. of the 1st

International Semantic Web Conference(ISWC),

pages333-347. Springer.

MSC: A Semantic Ranking for Hitting Results of

Matchmaking of Services, Proceedings of the 30th

Annual International Computer Software and

Applications Conference (COMPSAC'06)

O. Taha, T. Dhavalkumar, Al-Dabass D., 2006. Semantic-

Driven Matchmaking of Web Services Using Case-

Based Reasoning, IEEE International Conference on

Web Services.

Şenvar, M. and Bener, A., 2006, “Matchmaking of

Semantic Web Services Using Semantic- Distance

Information”, Lecture Notes in Computer Science by

Springer Verlag, ADVIS 2006, October, 18-20, İzmir,

Turkey.

ENASE 2007 - International Conference on Evaluation on Novel Approaches to Software Engineering

102