Intelligent Broker
A Knowledge based Approach for Semantic Web Services Discovery
Mohamed El Kholy and Ahmed Elfatatry
Institute of Graduate Studies & Research, Alexandria University, El horia street, Alexandria, Egypt
Keywords: Web Service, Service Broker, Semantic Web, Service Description, Web Service Description Language
(WSDL), Knowledge-based System.
Abstract: Ever since the introduction of the service oriented model of computing (SOA), service discovery has been
the major research challenge in SOA. Service consumers usually prefer to express their requirements
informally. Expressing requirements in such a way leads to difficulties in the matching procedure, and
hence results in poor matching results. In this paper, we present the concept of multi-level search as a
solution for matching informal expression of user requirements. In the suggested approach, intermediate
brokers receive service requests and suggest suitable services that match the given requests. We present a
mechanism by which an intelligent broker utilizes a knowledge based system to overcome the drawbacks of
syntactic and semantic discovery. The intelligent broker receives informal user requirements and performs
multi-level search. The search starts with key word search, then meaning search, and finally expert search.
If the keyword search fails to produce a proper matching, then, the search progresses to the following levels:
semantic, and then intelligent search. In this paper we argue that multi-level search could revive the dream
of automatic service discovery and present a detailed model for implementation.
1 INTRODUCTION
Service discovery is crucial for the success of
service oriented computing. However, it is still
posing a research challenge. Service discovery may
be performed manually or via automated
mechanism. In both cases, the searching interface
must be able to compare between the provided
capabilities and the required functionality. In this
paper, we present the concept of "multi-level search"
as a solution for discovery in service oriented
systems.
A major problem in service discovery is the
informal expression of user requirements. Unclear
ideas or ambiguous words in consumer requirements
lead to improper matchmaking results. The
consumer requirements are sent to an intermediate
broker that registers services from different
providers. A number of research efforts have
focused on enhancement of user requirements. One
solution is to use of common ontology to formally
describe user requirements (Zhang, 2011), (Baklouti,
2013). Such solution enforces the client to use
additional programs and hence affect the platform
independence of web services technology.
An alternative approach is to employ intelligent
brokers to improve service discovery. However,
existing web service brokers have failed to deliver
this promise (Zhang, 2011), (Zulkernine, 2011). A
significant amount of reasoning is performed by
such brokers to fulfill the required functionality.
However, existing brokers are still not intelligent
enough to deal with the complexity of informal user
requirements. Most brokers are capable of applying
keyword search and semantic search only (Zhang,
2011).
This paper introduces the “intelligent broker”
concept. In our solution, broker receives informal
specification of service functionality from the user
and then utilizes a knowledge based approach to
perform three levels of service discovery. The first
level is a syntax search that is based on key word
matching between the written user requirements and
the described service functionality. In case of no
matching, the user requirements are transformed to a
controlled English form and are provided with
available ontology. Semantic discovery takes place
as a second level. In case of semantic search failure,
the controlled user requirements are transformed
into predicate logic based search. Then, it is
39
El Kholy M. and Elfatatry A..
Intelligent Broker - A Knowledge based Approach for Semantic Web Services Discovery.
DOI: 10.5220/0005455300390044
In Proceedings of the 10th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE-2015), pages 39-44
ISBN: 978-989-758-100-7
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
converted to a knowledge based system supported
with rules to allow an intelligent search mechanism.
These multi levels of service discovery improve the
probability of accurate matching.
This paper is organized as follows. Section 2
discusses recent work on search brokers. Section 3
introduces the problems of syntactic and semantic
search. Section 4 presents our vision for intelligent
service discovery. Section 5 explains the proposed
method through a prototype of the intelligent broker.
2 THE BROKER ROLE IN
SERVICE DISCOVERY
2.1 Onto Broker
The Onto Broker architecture was introduced as a
semantic middleware (Paulheim, 2010). It is capable
of sharing meanings (semantics) of information and
capturing of complex relationships from
heterogeneous data sources. Onto Broker contains a
storage layer where ontologies from different
sources are stored. Such model requires high storage
and complex implementation of collaborating
servers. Also, customers have to integrate the Onto
Broker API into their own solutions to be connected
to onto broker.
2.2 Service Composition Broker
(Zhang) (2011) introduced a method to use the
broker to allow flexible semantic web service
composition. The broker first applies traditional
semantic search to find a service that fulfills the user
requirements. In case of matching failure the broker
employs an intelligent planner. The intelligent
planner generates sub tasks from the user
requirements and workflow knowledge. Then the
planner maps each task to a web service, and then
starts searching again. This process is repeated until
the user requirements are satisfied. However, this
approach utilizes the artificial intelligence to deal
with the user requirements not to preform service
discovery.
Most of the recent researches in this area depend
on the consumer side. They supply the consumer
with domain ontology or intelligent plans to improve
service provision. Our vision in this paper is more
general in that it accepts informal user requirements.
Then, the broker carries out a number of
transformations to search for suitable matches. This
guarantees faster and more accurate matching.
3 SERVICE DISCOVERY
PROBLEMS
3.1 Syntax Service Discovery
Syntax keyword discovery has proven to be
insufficient for powerful service discovery
(Žemlička, 2014). Web services are described by
XML file called Web Service Describing Language
(WSDL). WSDL allows developers to mainly
describe two essential parts of a developed service:
its functionality and how it can be invoked.
Matchmaking components at the broker side use the
functional descriptions to match users’ services
against their requirements. The XML Schema
Definition (XSD) language is used in the WSDL file
to express the structure of the message parts and data
types. XSD offers simple types (e.g., integer and
string). Hence, traditional WSDL file allows only
keyword searching and cannot allow semantic
reasoning. On one hand, matchmakers commonly
preprocess WSDL documents to extract terms that
may allow discoverers to find services relevant to
their requests. On the other hand, the user
requirement is supposed to be unclear and contain
ambiguous words. The weakness of service
descriptions or the unclearness of user requirements
may lead to two different problems. The first is the
matchmaking failure, which means that the user
request will have no reply at all. The second is to
have a long list of candidate services as a reply to
the user requirements. The user must analyze all
these services to find out the best service that exactly
fulfills his requirements.
3.2 Semantic Web Service Description
Semantic web services have limitations in the
discovery process due to the complex
implementations of their description methods.
Moreover, the semantic description methods suffer
from the lack of ontology standers among different
interacting parties (Rajagopal, 2006). The problem
related to semantic web service discovery has been
tackled in many directions. Semantic web service
description enhances web service discovery by using
non-ambiguous concept definitions from shared
ontologies (Zaremba, 2007). Thus the semantic
search overcomes the keyword search by utilization
interrelationships among data (Rajagopal, 2006).
The semantic data are machine understandable data,
and its relationships can be achieved through shared
ontologies (Ma, 2010). In this section we will focus
ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering
40
on two of the common semantic descriptions of web
which are WSMO-DF and OWL-S.
3.2.1 Web Service Modeling Ontology
Discovery Framework (WSMO-DF)
WSMO describes web services in a rich loosely
coupled semantic annotation. WSMO consists of
four main elements which are ontology, description,
goals, and mediator. The description element
describes the service according to the ontology
defined in the ontology element. Goals define the
user requirement. The mediator manages the
interaction between WSMO elements. WSMO-DF is
based on WSOA to offer an abstract description of
semantic services. WSMO-DF suggests two types of
Web service descriptions (Georgios, 2010). The
first is the abstract web service description which
defines a service in terms of its abstract
functionality. The second type is the concrete
service description which contains the details and
constrains needed for the service consumer. For
example, a hotel may offer an abstract service for
booking rooms. Then there are the detailed
descriptions and information about the service. For
instance, number of rooms, date, payment, etc. In
WSMO-DF the web service discovery mechanism
may use light or rich semantic descriptions. In the
light abstract description the abstract services are
represented as Complex Concepts (CCs). CCS map
similar services to the same class. For instance,
order class and search class. The rich semantic
discovery mechanism is the most fine-grained level,
where Web services are modeled in more detailed
specification including their state transitions.
3.2.2 Web Ontology Language for Services
(OWL-S)
OWL-S offers a conceptual model for semantically
annotating Web services. The model is based on four
ontologies, which are: Service, Service Profile (SP),
Service Process and Service Grounding. SP plays
the main role in service discovery while service
process and grounding provide information to use
the service. SP contains the main descriptive
information of the service including the service
name, and other useful information about the
provider. It describes the service functional
properties in terms of inputs and outputs (Niu,
2010). Profile based web service discovery is one of
the most commonly used semantic discovery
mechanisms. In such approach, the procedure of
matchmaking the service requests and service
description depends on that both are represented as
profile instances. These profiles include inputs and
outputs which are annotated with ontology. The
matchmaking is performed using semantic rule
formalisms. Listing 1 introduces the complex
concept and the service profiles in their semantic
service description.
The a1 service is classified in the Order class and
requires a passenger name, departure airport, arrival
airport, and flight number as input where the output
is e-Ticket which is sent to the user Email. a2 is a
more general service in the search class. It takes the
specification of a product. Then it searches for this
product according to its specifications. When a
matching product is found, the result (which is the
product name) is sent back to the user.
Complex Concepts
1) a1 ≡ Order Π ∀ passenger name.Name Π ∀ depart.Depart Π
∀ arrive.Arrive Π ∀ flighno.NO Π ∀ ticket.e-Ticket to { email}
2) a2 ≡ Search Π ∀ specification.Spesification Π ∀
product.Product
OWL-S Service Profile Instances
1) a1 : Orsder, (a1, Passenger Name): hasInput, (a1, Depart):
hasInput, (a1, Arrive): hasinput (a1, Flight No): hasinput, (a1,e-
Ticket) hasoutput
(a1, Email) : to
2) a2 : Search, (a2, specification) : hasInput,
(a2, product ): hasOutput
Listing 1: Semantic Web Service Description Examples.
The analytical study done in this work about
different semantic service description proved that
none of them is sufficient to ensure the process of
service discovery. Unfortunately, semantic web
service discovery suffers from high complexity and
the lack of standard ontologies. Furthermore
semantic search suffers from the absence of public
semantically annotated Web Services (Rajagopal,
2006). Moreover, Semantic web services should be
invoked and composed by clients who know
business process but are not aware of semantic
languages. The drawbacks of the syntactic and
semantic Web service discovery ensure us that web
services cannot depend on these kinds of search
(Žemlička, 2014). Web service needs more powerful
and intelligent searching mechanism to improve
service discovery.
4 INTELLIGENT SERVICE
DISCOVERY
The intelligent approach differs from other
IntelligentBroker-AKnowledgebasedApproachforSemanticWebServicesDiscovery
41
approaches in that it accepts traditional informal
consumer request. It does not enforce the service
consumer to apply any additional constraints to
formalize his requirements. The proposed system is
considered as a gateway between the syntax based
user requirement and the intelligent web world. This
approach is based on semantic web service
discovery in which all services must be registered at
a semantic matchmaker. The matchmaker includes
the semantic service description in a given ontology.
The first step in the proposed approach is to transfer
user requirements from informal English to formal
English language. We chose the Attempto controlled
English (ACE) to be our intermediate formal
language.
4.1 Attempto Controlled English
(ACE)
The (ACE) is a type of Controlled Natural Language
(CNL) which is used to control the user
requirements. ACE has been used for two reasons.
The first is that it has an easy structure as it is closer
to natural English. The second reason is because of
the possibility to be unambiguously translated into
predicate logic. Moreover ACE is used in
knowledge representation for the Semantic Web.
ACE is a formal language that contains only short
sentences. The structure of the sentence and its type
are well defined. These formal sentences can be
converted to first order predicate logic which
facilitates reasoning in semantic data.
4.2 Search Mechanism
The proposed system receives user request in XML
form. Then, three levels of search are applied to the
request. Figure 1 shows the three levels. At first,
traditional key word search is performed to find the
candidate service. If no matching occurs, the user
request is passed to the CNL unit which is supported
with English vocabulary. This unit introduces the
main elements of the sentence as subject, verb,
complements and adjuncts. This step solves the issue
of ambiguous sentence structure that could be
included in the user requirements. Then, formal
sentences are passed to ontology provider which
checks every word in the sentence and exchanges it
with the corresponding ontology word in the same
domain. As a result, the user requirements are
transferred from syntactic form to semantic form and
the second level of semantic search then takes place.
The search is performed at a semantic matchmaker
where the services are registered with their semantic
description. The matching between requested tasks
and the semantically described service is performed
as ontology reasoning at the semantic matchmaker.
The results of successful matches are passed back to
the requester. In case of no matching results, the
requester's requirements are transferred from the
ACE form to the predicate logic form. Then, the
logical requirements are passed to knowledge base
center which contains defined rules that improve the
service search. It also includes the history of the
registered services and knowledge on how to fulfill
the requester's tasks. The knowledge base includes
the history of the requester and his way to identify
his requirements. This knowledge base is connected
to an inference engine containing rules that support
intelligent search at the semantic matchmaker.
Figure 1: Three levels of service search.
These three levels of search syntactic, semantic
and intelligent increase the probability of matching
the required service. If no matches are found, a
failure report is sent to the requester advising him to
divide his requirements to individual tasks and try to
perform service composition to fulfill his main task.
5 PROPOSED BROKER
STRUCTURE
Figure 2 shows the structure of the intelligent multi-
level search broker. The proposed broker is based on
the semantic web services assumption where all the
services are semantically described. Service brokers
invoke the matchmaker to find service capabilities
that are included in the requester's requirements. The
proposed broker consists of five main units:
communicator, controller semantic search unit,
intelligent search unit and search engine unit.
A) Communicator: provides the basic interface to
communicant with services requesters. It also
monitors the results of each level of service search
ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering
42
until the matchmaking process is performed
successfully. It replies to the requester with the
service provider and the basic service description.
B) Controller: controls the execution of the broker.
It receives the request from the communicator
converting it to a task that has three levels of
processing. It manages the execution of these three
levels and their communication with the search
engine.
C) Semantic Search Unit: consists of two units
ACE converter and Ontology provider.
ACE Converter: is an intelligent engine supported
with English dictionary that is responsible for
converting the user requirements to control formal
English sentence. These ACE sentences are used to
perform ontology mapping in the semantic search.
Then, it is further converted to predicate logic
enabling intelligent search.
Ontology Provider: is responsible for matching
every word in the formal user sentence to its
corresponding ontology at the same domain.
D) Intelligent Search Unit: contains Knowledge-
base and services history. This unit stores service
requirements parameters represented in predicate
logic form. It also contains the history of invoked
services and the data to be transformed, including
inputs and outputs. The available knowledge is
controlled by rules supplied by the inference engine
of the knowledge-based system. Such rules and
knowledge allow the broker to perform intelligent
search.
E) Search Engine: The search engine calls the
matchmaker of semantic WS registry to find
appropriate services for a consumer's requirements.
The search engine performs traditional key search
and accepts data from the semantic search unit to
perform the semantic search at the matchmaker.
Figure 2: Proposed broker architecture.
6 A SEARCH CASE
Consider a case where a service consumer while
writing his code needs to invoke a service that
calculates the GPA for a specific student. He writes
his requirements using the words: college, subject,
and degrees. When these words are passed to the
intelligent broker, it will understand that the College
may have the same meaning as Faculty
and will start
representing the user requirement in predicate logic
in a form such as:
∀X: subject(X) course (X)
∀X: course (X) gpa (X, student)
In semantic repositories the data is saved in triples
called Resource Description Framework (RDF).
RDF consists of subject, predicate and object. In the
introduced example the data about a faculty is saved
in a semantic form in RDF triples as shown in
Figure (3). So, the intelligent broker can use the
predicate logic to search the web semantically.
Hence, the intelligent search can catch any triple that
gives the same meaning of the predicate logic
representation of the user requirements.
Figure 3: RDF graph representation in semantic web.
7 CONCLUSIONS
We introduced a mechanism for web service
discovery that overcomes the problems of keyword
search and search and the difficulties in semantic
matchmaking. The proposed intelligent broker
differs from other approaches in its searching
technique. Most of the existing approaches utilize
Natural language Processing (NLP) to enhance
semantic discovery. The proposed broker goes one
step further by utilizing ACE and predicate logic to
apply intelligent search. Predicate logic is a
promising technique to explore semantic web, since
the semantic web is composed of RDF triples based
on predicate logic.
REFERENCES
Blake, M. B., Cheung, W. K., Jäger, M. C., Wombacher,
IntelligentBroker-AKnowledgebasedApproachforSemanticWebServicesDiscovery
43
2007. WSC-07: evolving the web service challenge. In
IEEE Joint Conference on E-Commerce Technology
(9th CEC) and Enterprise Computing, E-Commerce
and E-Services.
Bleul, S., Weise, T., Geihs, K, 2007. Making a fast
semantic service composition system faster. In IEEE
Joint Conference on E-Commerce Technology (9th
CEC).
Bleul, S., Weise, T., Geihs, K.: The web service challenge:
a review on semantic web service composition. In:
Wagner, M., Hogrefe, D., Geihs, K., David, K. (eds.)
Service-Oriented Computing (SOC’2009), vol. 17.
European Association of Software, Science and
Technology (2009).
Brogi. A., Corfini, Popescu, 2008. Semantic-Based
Composition-Oriented Discovery of Web Service. In
ACM Transaction on Internet Technology 8(4), 19:1-
19:33.
Chiu I, Shu L., 2007. Biomimetic design through natural
language analysis to facilitate cross-domain analysis.
In (AI EDAM) Artificial Intelligence for Engineering
Design, Analysis and Manufacturing; 21:45–59.
Georgios Meditskos, Nick Bassiliades, 2010. Structural
and Role-Oriented Web Service Discovery with
Taxonomies in OWL-S. In IEEE Transactions on
Knowledge and Data Engineering 22(2), 278-290.
Heiko Paulheim, Florian Probst, 2010. Application
Integration on the User Interface Level: an Ontology-
Based Approach. In Data and Knowledge Engineering
Journal (DKE).
Huang, S., Wang, X., Zhou, 2005. Efficient web service
composition based on syntactical matching. In IEEE
International Conference on E-Technology, E-
Commerce, and E-Service (IEEE’05).
Luo, S., Xu, B., Yan, Y, 2010. An accumulated-QoS-first
search approach for semantic web service
composition. In IEEE International Conference on
Service-Oriented Computing and Applications
(SOCA’10).
Ma, H., Jiang, W., Hu, S., Huang, Z., Liu, Z, 2010. Two-
phase graph search algorithm for QoS-aware
automatic service composition. In IEEE International
Conference on Service-Oriented Computing and
Applications (SOCA’10).
Michal Žemlička and Jaroslav Král, 2014. Flexible
Business-oriented Service Interfaces in Information
Systems. In Proceeding of ENASE- 9
th
International
Conference on evaluation of Novel Approaches on
Software Engineering.
Muhammad Baqar Raza, Robert Harrison, 2011. Design,
Development & Implementation of Ontological
Knowledge based System for Automotive Assembly
Lines. In International Journal of Data Mining &
Knowledge Management Process.
Nabil Baklouti, Faten Fakhfakh, Bilel Gargouri, Mohamed
Jmaiel, 2013. OWL-LingS Editor A Tool for Semantic
Description of Linguistic Web Services. In CLOSER
2013, the 3rd International Conference on Cloud
Computing and Services Science. SCITEPRESS.
Niu, W., Chang, L., Wang, X., Han, X., Shi, 2010.
Semantic Web Service Discovery Based on Context
and Action Reasoning. In Pattern Recognition and
Artificial Intelligence 23(1), 65–71.
Rajagopal, S., Selvi, 2006. Semantic grad service
discovery approach using clustering of service
ontologies. In Proceedings of IEEE TENCON 2006,
pp. 1–4.
V. R. Benjamins, 1998. IBROW3 — An Intelligent
Brokering Service for Knowledge-Component Reuse
on the World Wide Web. In Proc. 11th Workshop
Knowledge Acquisition, Modeling, and Management
(KAW 98).
Yufeng Zhang, Hong Zhu, 2011. An Intelligent Broker
Approach to Semantics-based Service Composition. In
35th IEEE Annual Computer Software and
Applications Conference.
Zaremba, M., Vitvar, T., Moran, M., Brambilla, M., Ceri,
S., Cerizza, D., Della Valle, E., Facca, F. M.,
Tziviskou, 2007. Towards semantic interoperabilty:
in-depth comparison of two approaches to solving
semantic web service challenge mediation tasks. In
Ninth International Conference on Enterprise
Information Systems (ICEIS’07).
Zulkernine F. H., Martin, P., 2011. An Adaptive and
Intelligent SLA Negotiation System for Web Services.
In Services Computing, IEEE Transactions on
(Volume:4, Issue: 1), March.
ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering
44
