Enhancing Linguistic Web Service Description

with Non-functional NLP Properties

Nabil Baklouti

, Bilel Gargouri

and Mohamed Jmaiel

ReDCAD Laboratory, University of Sfax, Sfax, Tunisia

MIRACL Laboratory, University of Sfax, Sfax, Tunisia

Keywords:

Linguistic Web Service, Non-functional NLP Properties, Semantic Web, Ontology, OWL-S.

Abstract:

This paper deals with the enhancing of Linguistic Web Service (LingWS) description. It proposes an extension

for the OWL-S approach and a Natural Language Processing (NLP) domain ontology based on linguistic

standards. The proposed extension provides a classiﬁcation of the Non-functional NLP properties which

promotes the representation of their relationships. The extended OWL-S description is linked to the NLP

domain ontology to semantically annotate the LingWS properties.

1 INTRODUCTION

The lingware system engineering is a sub-domain of

the Software Engineering related to Natural Language

Processing (NLP). The development of this kind of

systems needs both several linguistic resources and

treatments. For this, researchers in NLP have re-

sorted to reuse existing lingware systems. These at-

tempts are based on the Web Service technology such

as (Ishida, 2006), (Tuﬁs et al., 2008), and (Baklouti

et al., 2010).

Today, many Linguistic Web Services (LingWS)

are published on the internet. They deal with various

applications such as Question/Response system and

Information Retrieval (Bramantoro et al., 2008). The

LingWS can be used like a simple Web Service or

integrated to a composite Web Service.

The Web Service discovery is the process of lo-

cating one or more related documents that describe

a particular Web service using the Web Services De-

scription Language (WSDL) (W3C, 2001) which rep-

resents a Service-Oriented Architecture (SOA) stan-

dard. Moreover, the SOA uses SOAP (Simple Ob-

ject Access Protocol) for the invocation and UDDI

(Universal Description Discovery and Integration) for

publishing Web Services. Unfortunately, these stan-

dards do not provide efﬁcient discovery results since

they are essentially focused on syntactical description

of Web Service. For example, the UDDI registry or-

ganizes the services according to deﬁned categories

without introducing semantic aspects. UDDI’s search

is also based on syntax like service’s name and it re-

lies on XML. The lack of semantics inside SOA stan-

dards powerfully inﬂuences the LingWS discovery. In

fact, there is a lack of NLP properties and their rela-

tions. For example, we should know which phenom-

ena are covered by a LingWS and what related ap-

proaches and resources are used. According to the

literature, several works suggested adding a seman-

tic wrapper or an NLP domain ontology for improv-

ing the description of LingWS. However, the Non-

Functional NLP properties and their relationships are

not considered in the above works.

Hayashi (Hayashi, 2011) has asserted that re-

searchers in the NLP ﬁeld have to develop an exter-

nal mechanism to semantically enrich the LingWS de-

scriptions in order to enhance its discovery.

The LingWS description should cover the Non-

Functional NLP properties and relationships between

them like the performed treatment type, the resource,

the analysis type, and so on. Unfortunately, the ex-

isting semantic approaches such as OWL-S (Martin

et al., 2004), WSMO (essi WSMO working group,

2004), and SAWSDL (Farrell and Lausen, 2007) are

unable to represent this kind of properties and their

relations.

In other domains, some extensions of OWL-S ap-

proach have been proposed such as (Aier et al., 2007)

and (Jean et al., 2010). These extensions aim to inte-

grate the quality standards of Web Services.

This paper proposes a semantic enrichment

of LingWS description by integrating the Non-

Functional NLP properties and their relations. For

this, we make an extension for OWL-S approach

439

Baklouti N., Gargouri B. and Jmaiel M..

Enhancing Linguistic Web Service Description with Non-functional NLP Properties.

DOI: 10.5220/0004080604390444

In Proceedings of the 7th International Conference on Software Paradigm Trends (ICSOFT-2012), pages 439-444

ISBN: 978-989-8565-19-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

which aims to both classify Non-functional NLP

properties and represent their relationships. In ad-

dition, we build a domain ontology using the NLP

norms to annotate the LingWS description. Thus, the

proposed solution is distinguished by allowing the de-

scription of the NLP speciﬁcities in a separate way

while highlighting linguistic links between them.

The remaining of this paper is structured as fol-

lows. Section 2 shows the attempts related to the

LingWS discovery problems. In section 3, we present

the Non-functional NLP properties. The proposed ex-

tension is given in section 4. Section 5 shows the NLP

domain ontology. In section 6, we provide a practi-

cal study to illustrate how the proposed extension can

represent the Non-functional NLP properties and their

relations. The last section concludes the paper.

2 RELATED WORK

We divide this section into two parts. We start with

presenting the relevant works in the NLP related to

the LingWS discovery. In the second part, we make a

comparative study of the semantic approaches.

In order to enhance the discovery of LingWS,

there are some relevant works which have proposed

to associate a wrapper around LingWS using se-

mantic technologies (i.e., OWL

and OWL-S) such

as (Ishida, 2006). It represents the LingWS Pro-

ﬁle which contains the LingWS Name, the LingWS

Type, a textual description, LingWS Status, and so on.

However, this proﬁle does not contain other relevant

Non-functional NLP properties and mainly their rela-

tions which may improve the LingWS discovery. An-

other issue is the absence of an ontology which rep-

resents both linguisic processing resources and their

Input/Output (I/O).

Klein and Potter (Klein and Potter, 2004) have

proposed an ontology for describing LingWS using

OWL-S approach. For our knowledge, it had explic-

itly stated the necessity of ontological foundation for

language infrastructure. Nevertheless, this proposi-

tion ignores taxonomies for both language resources

and abstract objects. In addition, the OWL-S is un-

able to both classify Non-functional NLP properties

and establish relationships between them. In terms of

linguistic resources interoperability, this proposition

does not take into account any NLP standard (e.g.,

LMF).

Hayashi (Hayashi, 2011) proposed an ”ontolo-

gization” of the Lexical Markup Framework (LMF

http://www.w3.org/2004/OWL/

http://www.lexicalmarkupframework.org/

This work does not represent the LingWS I/O

which are important for ensuring LingWS discovery.

Hayashi et al. (Hayashi et al., 2008),(Hayashi, 2011)

used SAWSDL to annotate LingWS description. Nev-

ertheless, SAWSDL cannot represent the details of

the NLP knowledge. Moreover, Hayashi in (Hayashi,

2011) has asserted that researchers in the NLP ﬁeld

have to develop a mechanism for discovering allow-

ing the semantic enrichment of LingWS description.

To conclude, the LingWS description should be

augmented with Non-functional properties and their

relationships which can enhance the discovery task.

In addition, the NLP domain ontology should be more

expressive in terms of NLP speciﬁcities.

In order to overcome the WSDL semantic

lack, various approaches have been proposed such

as OWL-S (Martin et al., 2004), WSMO (essi

WSMO working group, 2004), and SAWSDL (Farrell

and Lausen, 2007).

The OWL-S approach is built inside the Web Ser-

vices. It proposes an ontology of services motivated

by the need to provide three elements: The Proﬁle

which is used to announce the service. It contains

the I/O, the preconditions, the results, and the service

category of Web Service. The Process which con-

tains I/O, preconditions, results, and the behaviour of

the service (data and control ﬂow), and the third el-

ement is the Grounding which provides the details

(e.g, protocol, address) to invoke the services.

The Web Service Modelling Ontology (WSMO)

provides a framework for semantic descriptions of

Web Services and acts as a meta-model for such Ser-

vices based on the Meta Object Facility (MOF)

. Se-

mantic service descriptions, according to the WSMO

model, can be deﬁned using Web Service Model-

ing Language (WSML

). It consists of four elements

deemed necessary to support Semantic Web services:

Ontologies, Goals, Web Services, and Mediators.

Semantic Annotations for WSDL and XML

Schema (SAWSDL) deﬁnes a new name-space called

”sawsdl”. There are three extensions for it : mod-

elreference which associates an XML Schema or a

WSDL component to an ontology concept. The other

two extensions are liftingSchemaMapping and low-

eringSchemaMapping which promote the mapping

between the semantic data and the XML elements.

We use some criteria to compare the above ap-

proaches: For SAWSDL, it deals with the discov-

ery and the automatic invocation of Web Services but

not for the composition. With SAWSDL, we can

use any type of ontology (e.g, OWL, WSML) while

OWL-S supports only OWL ontologies and WSMO

http://www.omg.org/mof/

http://www.wsmo.org/TR/d16/d16.1/v0.21/

ICSOFT2012-7thInternationalConferenceonSoftwareParadigmTrends

440

the WSML ontologies. Otherwise, the SAWSDL does

not promote the deﬁnition of Non-Functional prop-

erties. In terms of tools, OWL-S provides several

tools such as editor, matcher, and composer. The

WSMO tools are more difﬁcult to develop because

they are based on WSML which is never used in the

past nevertheless OWL-S and SAWSDL rely on RDF

and XML.

We clearly note that OWL-S approach provides

more beneﬁts in terms of both ontology, tools, Non-

Functional properties, and composition. Thereafter,

OWL-S seems to be the best approach for describing

LingWS.

3 NON-FUNCTIONAL NLP

PROPERTIES

Several types of information need to be modelled

within the LingWS description. For this, we present

in the following some Non-functional NLP proper-

ties. We classify these properties by processing level:

- Lexical Level: It is characterized by the use of

lexical linguistic resources, lexical approaches,

lexical formalisms, lexical analysis types and so

on.

For example, the developer can choose its

LingWS according to the used lexical analysis

such as the thematic (It proposes a large concep-

tual category in which the user can navigate for

ﬁnding the suitable word), the structural (It helps

the writer in the structure choice) or the syntag-

matic one (It is a statement element regrouped into

sub-parties with an internal structure and a coher-

ent unit).

- Morphological Level: It contains various fea-

tures such as morphological phenomena, morpho-

logical formalisms, and approaches. As an exam-

ple, there are the Linguistic approach (It segments

a text to elementary units which have a linguistic

knowledge attached: grammatical category , gen-

der, number , time, person and so on.), the Statis-

tic approach (The analysis starts by splitting sen-

tences into words. Then, a cost is attributed to

each bi-gramme according to the calculated ap-

parition frequency in a corpus. Finally, the solu-

tion which has the lower cost is chosen like the

best probable.), and the Hybrid approach (It com-

bines linguistic and statistic criteria. It extracts the

relevant terms from both text statistic analysis and

linguistic ﬁltering of the candidate terms. It pro-

duces a sorted list of the most representative terms

for a speciﬁc domain.).

We show the requirement of the approach kind by

the following example : When a developer has

the intention to build a morphological application

which uses a Linguistic approach, so he has to

take a way all LingWS using other kinds of ap-

proaches.

- Syntactic Level: Different speciﬁcities can char-

acterize a syntactic LingWS such as syntactic

phenomena, syntactic analysis, and syntactic for-

malisms. As an example, for the analysis type

we can mention: Top-down analysis (The anal-

ysis begins from the start symbol called axiom

and try to rebuild the derivation tree by a pre-

ﬁxed left-right course.), Bottom-Up analysis (It

factorizes the word by picking out or recogniz-

ing the right parts of production until ﬁnd out the

axiom.), Profound analysis (It produces a formal

representation of the sentences, under a syntactic

tree form.), Surface or Chunking (It identiﬁes the

components limits i.e., Nominal Group (NG) and

Verbal Group (VG)) and Structural analysis (It is

based on a set of rules for deﬁning associations

between words in order to construct sentences.).

When an application treats two phenomena (e.g.,

Accord and Anaphora) using two formalisms

(e.g., Uniﬁcation Grammar and Resolution Algo-

rithm), so the developer has to choose the suitable

LingWS according to the used formalism for each

retained phenomenon.

- Semantic Level: To develop a semantic applica-

tion, we can choose some NLP properties such

as how to represent knowledge, the semantic for-

malisms, the semantic phenomena, and the se-

mantic resources. The used resource is a rele-

vant information. Indeed, if a developer wants to

compose an application which needs a Wordnet

resource then he has to eliminate LingWS using

LMF resource for example.

After presenting the Non-Functional NLP properties,

we observe that a LingWS description should contain

this kind of properties for enhancing LingWS discov-

ery.

4 EXTENSION OF OWL-S

4.1 Drawbacks of OWL-S

The Proﬁle promotes the description of both Func-

tional and Non-Functional properties of Web Ser-

vices. It is associated to a set of ServiceParameter

which allows the annotation of services by couples

(criteria, value). However, the latter cannot be used to

EnhancingLinguisticWebServiceDescriptionwithNon-functionalNLPProperties

441

Figure 1: Extension of OWL-S for integrating Non-functional NLP properties.

represent relationships between Non-functional NLP

properties. In fact, it is not possible to: (1) Use such

criterion to classify the Non-functional NLP proper-

ties, so using OWL-S model and for one LingWS, we

cannot obtain both processing level and its linguis-

tic properties such as phenomena, resources, and for-

malisms. (2) Remove ambiguity: If a LingWS covers

many phenomena and each one is treated by one for-

malism, so we cannot keep these relations. (1) and (2)

prove that an extension of OWL-S can be very useful

to further exploit the speciﬁed Non-functional NLP

properties of LingWS.

4.2 Proposed Extension

The OWL-S extension is presented by Figure 1. As it

is shown by this ﬁgure, the OWL-S extension is based

on the specialization of ‘ServiceParameter‘ class by

one class namely ‘ServiceProcessing Level‘.

The main elements of the proposed extension are:

• ServiceProcessing Level: It represents the pro-

cessing level of the LingWS. We have essentially

four processing levels which are Syntactic, Lexi-

cal, Morphological, and Semantic. Each process-

ing level is characterized by both its resources

(e.g., dictionaries, tree bank, corpus) and phenom-

ena (e.g., ellipsis, anaphora, accord). For this rea-

son, we add respectively the Resource and the Lin-

guisticPhenomenon classes.

• LinguisticPhenomenon: It has the ‘reﬁned into‘

relation, since each phenomenon has its sub-

Phenomena. For example, for the ellipsis phe-

nomenon, we can ﬁnd the nominal ellipsis (the

omission of the essential part of a nominal phrase:

the head) and an ellipsis of a whole phrase (e.g.,

subject ellipsis, verb ellipsis, both verb and com-

plement ellipsis). The LinguisticPhenomenon has

also other relations with other classes:

– Approach: the treatment approach of a such

phenomenon.

– LinguisticFormalism: represents the formalism

(e.g., HPSG and LFG for syntactic Grammars).

It supports the phenomenon determination us-

ing an Analysis Type (e.g., Top-Down, Bottom-

Up).

5 NLP DOMAIN ONTOLOGY

It is essential to develop an NLP domain ontology

to annotate the elements of the LingWS description

(e.g, I/O and the recently added properties). Thus, the

proposed extension and the NLP ontology should be

linked using pointers (e.g, sParameter) in order to give

an expressive semantic description. Unfortunately,

the existing ontologies are incomplete and not spe-

cialized to improve the LingWS discovery (Hayashi

ICSOFT2012-7thInternationalConferenceonSoftwareParadigmTrends

442

Figure 2: Extract from NLP domain ontology.

et al., 2008). Hence, we have developed a new on-

tology which both promotes linguistic resource tax-

onomies and takes advantages of ISO standards pro-

posed in the area of lexical resources construction,

namely LMF (ISO 24613) and DCR (ISO 12620).

We have used these standards to identify concepts

and data-properties (i.e, relations). The developed on-

tology is extensible, so we can add other resources

(i.e., data resource and processing resource). We used

OWL

as description language and Prot

as the

main tool for the ontology construction. Figure 2

shows an extract of the developed ontology dealing

with the Morphological and the Syntactic levels.

6 DEMONSTRATION

To consolidate our solution, we use a service li-

brary available in our laboratory which contains many

LingWS then, we take one for applying the OWL-S

extension using the NLP domain ontology. Currently,

our library contains about 40 LingWS which can be

expanded by other. The available LingWS cover some

languages : Arabic, French and English. We obtained

these LingWS from both (Baklouti et al., 2010) and

from some opensource tools like OPEN-NLP

, NLP-

LIB, classiﬁer4j, standford

, and so on. The major

of these tools are used by the known platforms (e.g,

GATE and UIMA).

We choose the ‘Syntactic Parser‘ as an example

of LingWS for making a practical study. The latter

promotes many tasks such as it performs context-free

syntax analysis, it guides context-sensitive analysis,

and it attempts error correction.

Now, we present the ‘Syntactic Parser‘ proﬁle us-

ing the proposed extension.

For this, we use OWL-S Editor

which is a plug-

in for prot

e. It is the most used tool for OWL-S,

http://www.w3.org/TR/2009/REC-owl2-overview-

20091027/

http://protege.stanford.edu/

http://incubator.apache.org/opennlp/

http://nlp.stanford.edu/links/statnlp.html

http://owlseditor.semwebcentral.org/index.shtml

since it performs several tasks such as edition, compo-

sition, and execution. Moreover, it is able to both load

WSDL ﬁles of existing LingWS, generate an OWL-S

‘skeleton‘ using wsdl2owls class, annotate generated

description, and execute Semantic LingWS.

Figure 3 shows how we represent the Non-

Functional NLP properties as well as the relation-

ships between them. This representation is not pos-

sible using the initial OWL-S proﬁle. To do this,

we added some ‘Object Properties‘ like ‘reﬁned into‘,

‘treated By‘, and so on.

As it is indicated by Figure 3, the ‘Syntac-

tic Parser‘ LingWS treats ‘Ellipsis‘ as ‘Linguis-

ticPhenomenon‘ which is reﬁned into two sub-

phenomena: ‘Nominal Ellipsis‘ and ‘Verb Ellipsis‘.

The ﬁrst one is resolved using ‘Semantic Approach‘

and the second one by an ‘Hybrid Approach‘ as ‘Ap-

proach‘.

Figure 3: OWL-S extension code of ‘Syntactic Parser‘

Figure 3 also shows the links between the Non-

Functional NLP properties and the NLP domain on-

tology by respectively ‘sPhenomenon‘ for ‘Linguis-

ticPhenomenon‘ and ‘sApproach‘ for ‘Approach‘.

This ﬁgure does not contain the initial OWL-S el-

ements such as ‘serviceName‘, ‘textDescription‘ but

only the recently added NLP properties (i.e, ‘Linguis-

ticPhenomenon‘, ‘Approach‘).

This section clearly prove how the proposed ex-

tension can both represent the Non-functional NLP

properties and establish relationships between them.

EnhancingLinguisticWebServiceDescriptionwithNon-functionalNLPProperties

443

7 CONCLUSIONS AND FUTURE

WORK

We provided a solution to the problems related to the

lack of semantic within the LingWS description. In-

deed, we proposed an OWL-S extension for integrat-

ing Non-functional NLP properties and relations be-

tween them, since they can semantically enhance the

LingWS descriptions. Besides, we developed an NLP

domain ontology using NLP ISO standards.

The demonstration that we made proves the ef-

fectiveness of the OWL-S extension. Actually, the

LingWS description is more meaningful, so it pro-

motes advanced research (i.e, many combinations)

knowing also that each property can be accessible

alone.

For the future, we plan to cover other NLP prop-

erties mainly for the non-Latin languages. Then, we

will check out the LingWS discovery by deﬁning an

appropriate matching algorithm. Finally, we will ad-

dress the LingWS composition.

ACKNOWLEDGEMENTS

A Part of this work was done during a SIMBAD sup-

ported internship at the TELECOM SudParis, Evry

France. For this, the authors thank Samir Tata for its

support.

REFERENCES

Aier, S., Offermann, P., Schonherr, M., and Schropfer, C.

(2007). Implementing non-functional service descrip-

tions in soas. In Trends in Enterprise Application Ar-

chitecture Lecture Notes in Computer Science, volume

4473/2007, pages 40–53.

Baklouti, N., Bouaziz, S., Gargouri, B., Aloulou, C., and

Jmaiel, M. (2010). Towards the reuse of lingware sys-

tems: A proposed approach with a practical experi-

ment. In Proceedings of the International Conference

on Information Integration and Web-based Applica-

tion and Services (iiWAS2010) ACM, pages 566–572,

Paris-France.

Bramantoro, A., Tanaka, M., Murakami, Y., Schafer, U.,

and Ishida, T. (2008). A hybrid integrated architec-

ture for language service composition. In IEEE Inter-

national Conference on Web Services.

essi WSMO working group (2004). Web service

modeling ontology. http://www.wsmo.org/.

http://www.wsmo.org/.

Farrell, J. and Lausen, H. (2007). Semantic annotations for

wsdl and xml schema. http://www.w3.org/TR/sawsdl.

http://www.w3.org/TR/sawsdl.

Hayashi, Y. (2011). Prospects for an ontology-grounded

language service infrastructure. In Proceedings of

Workshop on Language Resources, Technology and

Services in the Sharing Paradigm, page 17, Chiang

Mai, Thailand.

Hayashi, Y., Narawa, C., Monachini, M., Soria, C., and Cal-

zolari, N. (2008). Ontologizing lexicon access func-

tions based on an lmf-based lexicon taxonomy. In

LREC, Marrakech, Morocco. European Language Re-

sources Association (ELRA).

Ishida, T. (2006). Language grid: An infrastructure for in-

tercultural collaboration. In proceedings of the Inter-

national Symposium on Applications and the Internet

(SAINT06) IEEE, pages 96–100.

Jean, S., Losavio, F., Matteo, A., and L

evy, N. (2010). An

extension of owl-s with quality standards. In Pro-

ceedings of the Fourth IEEE International Conference

on Research Challenges in Information Science, pages

483–494.

Klein, E. and Potter, S. (2004). An ontology for nlp ser-

vices. In Proc. of LREC Workshop on a Registry of

Linguistic Data Categories within an Integrated Lan-

guage Resource Repository Area.

Martin, D., Paolucci, M., McIlraith, S., Burstein, M., Mc-

Dermott, D., McGuinness, D., Parsia, B., Payne, T.,

Sabou, M., Solanki, M., Srinivasan, N., and Sycara,

K. (2004). Bringing semantics to web services: The

owl-s approach. In International Workshop on Seman-

tic Web Services and Web Process Composition.

Tuﬁs, D., Ion, R., Ceausu, A., and Stefnescu, D. (2008).

Racais linguistic web services. In Proceedings

of the Sixth International Language Resources and

Evaluation (LREC08), Marrakech, Morocco. Eu-

ropean Language Resources Association (ELRA).

http://www.lrec-conf.org/proceedings/lrec2008/.

W3C (2001). Web services description lan-

guage (wsdl). http://www.w3.org/TR/wsdl/.

http://www.w3.org/TR/wsdl/.

ICSOFT2012-7thInternationalConferenceonSoftwareParadigmTrends

444