Ontology-Based Service Discovery Front-End Interface for GloServ
ABSTRACT This paper describes an ontology-based service discovery front-end interface for GloServ. GloServ is a service discovery engine,
which is an ontology-based distributed service discovery system that allows sophisticated querying of services. The working
implementation of the front-end interface demonstrates how GloServ can be used for different types of web service discovery.
The front-end generates a search form from the service class ontology. It also allows multiple services to be queried for
in a single search by generating cascaded forms for combined service queries. It then converts the input to a GloServ query
and displays the results to the user in a coherent manner. The use cases that are demonstrated with this implementation are
service discovery for location-based services, tagged services and collaborative search with other users.
- SourceAvailable from: Ying Zou[Show abstract] [Hide abstract]
ABSTRACT: In Service Oriented Architecture (SOA), service composition integrates existing services to fulfill specific tasks using a set of standards and tools. However, current service composition techniques and tools are mainly designed for SOA professionals. It becomes challenging for end-users without sufficient service composition skills to compose services. In this paper, we propose a framework that supports end-users to dynamically compose and personalize services to meet their own context. Instead of requiring end-users to specify detailed steps in the composition, our framework only requires the end- users specify the goals of their desired activities using a few keywords to generate a task list. To organize the task list, we analyze the historical usage data and recover the control flows among the tasks in the task list. We also mine the task usage pattern from the historical usage data to recommend new services. A prototype is designed and developed as a proof of concept to demonstrate that our approach enables end-users to discover and compose services easily.The Smart Internet - Current Research and Future Applications; 01/2010
Conference Paper: Task-driven service discovery and selection.[Show abstract] [Hide abstract]
ABSTRACT: Services are becoming more and more widely used. When designing interactive applications based on services one important issue is how to identify those services most relevant for the application functionalities. The proposed approach takes as input a task model, which includes the user's view of the interactive system, and an ontology capturing the application domain, and automatically discovers a set of ordered service descriptions for each system task of the model. The discovered descriptions can be used in order to invoke a particular service operation that fulfils a task's required functionality. In this way, the whole application functionality can be realized by a set of service operations without writing a single line of code. As a result, the application development time is significantly reduced and it is possible to complete the development of interactive front-ends by integrating our solution in existing model-based HCI approaches.Proceedings of the International Conference on Advanced Visual Interfaces, AVI 2010, Roma, Italy, May 26-28, 2010; 01/2010
Conference Paper: An automatic approach for ontology-driven service composition[Show abstract] [Hide abstract]
ABSTRACT: Current service composition techniques and tools are mainly designed for use by service oriented architecture (SOA) professionals to solve business problems. This focus on SOA professionals creates challenges for the non-expert users, with limited SOA knowledge, who try to integrate SOA solutions into their online experience. To shelter non-expert users from the complexity of service composition, we propose an approach which automatically composes a service on the fly to meet the situational needs of a user. We present a tag-based service description schema which allows non-expert users to easily understand the description of services and add their own descriptions using descriptive tags. Instead of specifying the detailed steps for composing a service, a non-expert user would specify the goal of their desired activities using a set of keywords then our approach can automatically identify the relevant services to achieve the goal at run-time. A prototype is developed as a proof of concept. We conduct a case study to compare the performance of our approach in automatic service composition with a baseline approach which consists of the manual process of searching for services using keywords. The case study shows that our approach can achieve higher precision and recall than the baseline approach.Service-Oriented Computing and Applications (SOCA), 2009 IEEE International Conference on; 02/2009
Ontology-based Service Discovery Front-end Interface
Knarig Arabshian, Christian Dickmann and Henning Schulzrinne
Department of Computer Science
Columbia University, New York NY 10027, USA
Abstract. This paper describes an ontology-based service discovery front-end
based distributed service discovery system that allows sophisticated querying of
services. The working implementation of the front-end interface demonstrates
how GloServ can be used for different types of web service discovery. The front-
end generates a search form from the service class ontology. It also allows mul-
tiple services to be queried for in a single search by generating cascaded forms
for combined service queries. It then converts the input to a GloServ query and
displays the results to the user in a coherent manner. The use cases that are
demonstrated with this implementation are service discovery for location-based
services, tagged services and collaborative search with other users.
Keywords: user interface, service discovery, ontologies, OWL, CAN, peer-to-
This paper describes an ontology-based service discovery front-end user inter-
face to GloServ   . GloServ is an ontology-based global service
discovery system. It uses the Web Ontology Language Description Logic (OWL
DL)  to classify services in an ontology and map knowledge obtained by the
ontology onto a hybrid hierarchical peer-to-peer network. It operates in wide
as well as local area networks and supports a large range of services that are
aggregated and classified in ontologies. A partial list of these services include:
events-based, physical location-based, communication, e-commerce or web ser-
GloServ provides a generic back-end service discovery framework for dif-
ferent front-end systems to interact with it. The contribution of this paper is
a working implementation of a web-based front-end which demonstrates how
GloServ can be used for different types of web service discovery. The use
cases that are demonstrated with this implementation are service discovery for
location-based services, tagged services and collaborative search with other
2LATEX style file for Lecture Notes in Computer Science – documentation
In order to build a web-based front-end to GloServ, the following prob-
lems need to be addressed: 1) downloading the service class ontologies from
the GloServ back-end; 2) generating a search form from the service class on-
tology files; 3) allowing multiple services to be queried for in a single search
by generating cascaded forms for combined service queries; 4) converting the
user input to a GloServ query; 5) displaying the results to the user in a coherent
Below, we describe the solution of the design and implementation of the
web-based front-end system. We begin by giving an overview of the back-end
GloServ service discovery system in Section 2. In Section 3 we motivate the
need for a web-based front-end by describing a few use cases. Sections 4 and
5 describe the front-end system and its implementation, respectively. Finally we
discuss related work in service discovery in Section 6 and conclude in Section 7.
2 Overview of GloServ
This section gives a brief overview of GloServ, the back-end service discovery
architecture. For further details on the design, implementation and evaluation of
GloServ, we encourage the reader to refer to   .
One of the main components of GloServ is the service classification ontol-
ogy. At a high-level, services are classified in a pure hierarchy where each ser-
vice class is disjoint from the other. Hence, service instances will only be classi-
fied within one of the branches. At the lower levels of the ontology, classes may
have relationships with other classes and a pure hierarchy is not maintained. The
upper hierarchical ontology which defines high-level services is mapped onto a
hierarchical network and the low-level ontologies are mapped to a peer-to-peer
Another component of GloServ is the back-end hybrid hierarchical peer-to-
peer service discovery network. The high-level hierarchical ontology maps to a
physical hierarchical network. Each service class ontology maps to the Content
Addressable Network (CAN)  which is a peer-to-peer network architecture.
Since GloServ achieves load distribution, fast query and update processing time,
while maintaining reliability, we have elected this as the underlying service dis-
GloServers maintain three types of information: a service classification on-
tology, a thesaurus ontology and, if part of a peer-to-peer network, a CAN
lookup table. The high-level service classification ontology is not prone to fre-
quent changes and thus can be distributed and cached across the GloServ hier-
archical network. Each high-level service will have a set of properties that are
inherited by all of its children. As the subclasses are constructed, the proper-
LATEX style file for Lecture Notes in Computer Science – documentation3
ties become specific to the particular service type. The thesaurus ontology maps
synonymous terms of each service to the actual service term within the sys-
tem. Figure 1 gives an overview of the GloServ architecture generated from the
Restaurant ontology and how queries are routed to the correct servers.
Fig.1. Finding servers in GloServ
At the lower levels, maintaining a purely hierarchical ontology structure be-
comes difficult as classes tend to overlap. Thus, in order to efficiently distribute
service instances according to similar content, servers that hold information on
similar classes are distributed in a peer-to-peer network. The CAN peer-to-peer
architecture distributes classes with similar content and is generated as a net-
work of n-level overlays, where n is the number of subclasses nested within
the main class. The first CAN overlay is a d-dimensional network which has
the first level of subclasses of the Restaurant class. The number of dimensions
is determined by the maximum number of nodes which can be added into the
CAN. This is estimated to be (log2n)/2 to ensure that the number of query hops
4LATEX style file for Lecture Notes in Computer Science – documentation
are O(log2n). Services are represented as instances of the service classes and
reside in the CAN servers.
A user initially contacts a GloServ user agent and enters a service name.
The initial GloServer is found after following the steps outlined in Figure 1.
Since each hierarchical node handles a class which is disjoint from its siblings,
the query is routed down only one branch reducing the query hops considerably.
Once the correct GloServer is contacted, the user agent obtains the ontology
pertaining to that service class. The interface to the user can either be human-
centric or automated, depending on the implementation. In either case, a query
is formed and sent to the GloServer. The query is a first order predicate logic
statement that contains restrictions on various properties such as: (hasLocation
some NYC) and (hasCuisine some (Korean or Chinese)). The server han-
dling that service class creates a class with this query restriction and classifies it
in its ontology. Since the subclasses of the Restaurant class are restricted by lo-
cation, the query class gets classified as a subclass of the NYCRestaurant class.
The query is then forwarded to the nodes that handle NYCRestaurant classes.
When a node is found, the query class is classified again. Since the NYCRestau-
rant class has subclasses that have cuisine restrictions, when the reasoner clas-
sifies the query class, it becomes a subclass of NYCRestaurant and a superclass
of KoreanNYCRestaurant and ChineseNYCRestaurant classes. The classifica-
tion indicates that the query must be routed to the servers handling Chinese and
Korean restaurants. In order to route the query within a CAN, the query needs
to reduce to a dimension and key. The dimension and key values assigned to
each of these classes are used during the CAN network generation to convert
the ontology class to a < dimension,key > pair.
GloServ allows ontology-based querying with key word matching. Since the
service ontology may not capture all parts of a particular service, we allow each
ontology to have a keyWords property which service providers can populate
with keyword terms that describe the specifics of their service. The ontology-
based query is executed and the remaining results are refined by matching the
keywords entered. Gloserv also allows a combination of queries across different
service classes. Thus, one can search for services with a common property, such
as location, in a single search. An example of this is querying for a restaurant
and a nearby movie theater.
The motivation to create a web-based front-end to GloServ is to demonstrate
that GloServ can be easily plugged into a Web Service Architecture  and per-
form interesting service discovery searches. The Web Service Architecture may
LATEX style file for Lecture Notes in Computer Science – documentation5
have different types of service discovery systems such as centralized registries
as in UDDI , text-based search engines as in Google or peer-to-peer archi-
tectures such as GloServ. Using GloServ for web service discovery shows that
service discovery can be done using a richer description language as well as
within a distributed architecture. We demonstrate that a web service front-end
can be implemented which discovers services using GloServ. However, we have
not built a front-end which interacts with the service providers directly using
SOAP and HTTP as this is not applicable to the service discovery phase and is
beyond the scope of this paper. Below, we describe three use cases that a web-
based front-end to GloServ can be used for: location-based service discovery,
searching for tagged services and collaboratively searching for services.
Consider location-based services such as restaurants, theaters, and traffic. Imag-
ine a couple is planning an evening out in New York City and would like to
find a seafood restaurant near the waterfront in Manhattan that also has a the-
ater nearby playing an action movie. In order to find the quickest route to the
restaurant, the closest restaurant which has the least congested route needs to be
found. It should also be located near a movie theater which is featuring action
movies with showtimes that are some time after dinner is over.
To perform the above search, location data must be integrated across the
different domains. First, seafood restaurants are searched for in New York City
which have an additional attribute of being located near a waterfront. The loca-
tion of these restaurants are then fed into a traffic service network which returns
the best route to the restaurant. The locations of these restaurants are chosen
and then fed into the theater service network along with attributes for movie and
showtimes. The results retrieved will be all the restaurants and nearby theaters
which are located in places that offer the least congested route.
Currently, in order to perform such a search, a user must issue multiple
queries to restaurant, theater and traffic sites in order to determine the best
restaurant and theater to go to. Since data for each of these services is repre-
sented in different formats, there is no way to integrate the data while com-
posing services. The use of ontologies would greatly aid in performing these
types of complicated queries. Furthermore, if we take into account all types of
location-based services, global distribution of data becomes necessary. Thus, we
have chosen to use GloServ as the underlying service discovery system since it
provides an ontology-based service search in addition to global distribution of
In this use case, three service classes are queried: Restaurant, Theater, and
Traffic Reports. These classes all share a common property, namely, location.
6LATEX style file for Lecture Notes in Computer Science – documentation
Since an ontology is useful in mapping semantically equivalent concepts to one
another a Location ontology can be defined and shared across all the different
service domains. The front-end downloads ontologies for the multiple service
classes by providing links to the service forms. It then exhibits three cascading
forms that appear to the user, one for each service class with the shared location
field appearing once. The results are displayed in a Google map.
Another example of a use case for a web-based front-end is searching for ser-
vices and their tags. Currently, tagging is becoming a common phenomenon in
various web services such as Flickr  and de.licio.us . We define a tag as an
many users like to read and give feedback pertaining to services they will use or
have used. This review system can be deemed as a Tag service class which has
any number of different rating services such as Zagat  for restaurants, Better
Business Bureau for businesses, or regular user reviews. Given this type of ser-
vice composition, one can search for any service which has a particular rating
and give feedback for this service as well. A Tag service class is defined with its
own service classification ontology and inserted into the GloServ network.
The relationships between GloServ service classes and their corresponding
tag services can be determined using an ontology mapper. For example, tags
from the Zagat and NY Times classes match the Restaurant service class for
restaurant ratings and the NY Times class also matches the Theater class for
movie ratings. Flickr and Weather tag classes match the SightSeeing service
class because each of these can tag various travel sites such as travel brochures
or tourist spots. The front-end establishes the relationship table of a given ser-
vice class and its corresponding tagging service by using an ontology mapper.
Figure 2 shows the relationships between Tag services and Location-based
Services. Front-end web services handling sightseeing or restaurant services
know about the tagging providers that contribute tags to their services via this
relationship ontology. In most cases, these tags will be contributed by third party
Searching for restaurants and theaters with specific reviews becomes simple
with thefront-end we havedesigned. The userqueries forRestaurantandRating
by choosing these service links and entering data into the restaurant and rating
forms. Users can also add tags to a service by providing the service URN and
the reviews in the rating form.
LATEX style file for Lecture Notes in Computer Science – documentation7
Fig.2. Relationship ontology of the Tag class and the Location-based Services class
An interesting application of GloServ is letting users collaborate in a single
search. Imagine two mobile users who decide on a last minute dinner meeting.
They would like to perform the search together but one of them is sitting in a
meeting and is unable to talk. Thus, her friend invites her to collaborate with her
on a service discovery search through GloServ. The invitation is sent via email
or text message which includes a link to the GloServ collaborative interface. As
userAclickson herpreferences,thevalues changeonuserB’sfront-end aswell,
by synchronizing the interfaces. The values entered by both users is converted
to a regular GloServ query and issued to GloServ.
The front-end is a web server that runs Apache and PHP  (version 5.1.x).
It allows users to register and query for location-based services. Currently, the
service classes that are supported are for Restaurant, Theater, Weather and Tag-
ging services. The interface provides links for each service class. When the
user clicks on a service class name, a form is generated for that service class.
The query results are displayed in a Google map as well as in a list. The over-
all GloServ front-end can be seen in Figure 3. This section describes how the
front-end operates in order to perform location-based service discovery using
8LATEX style file for Lecture Notes in Computer Science – documentation
Fig.3. GloServ Front-End
4.1Generating a Search Form
The first step that a front-end needs to accomplish is to download the correct ser-
vice ontology files from the GloServ back-end. Depending on what the web ser-
vice is, it contacts the GloServ back-end by submitting the service class name.
This query is routed through the GloServ hierarchy as described in Section 2
and the ontology is returned to the front-end server. The front-end caches these
ontology files as well as the hostnames for the CAN super nodes of the service
classes and refreshes these periodically. In this way, the query does not have to
route through the hierarchical servers but can go directly to the CAN level.
The front-end parses the ontology and displays each property as a field in
the form. Each property is annotated with a label property which is used for
the graphical user interface. The ontology parser parses out each property’s la-
bel and displays it as the field’s label in the form. The form field is determined
by the property’s range. If the range is an object property, all the classes in
the range are parsed and displayed as a drop-down menu. For datatype proper-
ties, the field is a regular text box. For example, the hasCuisine property has
annotation properties label.us=Cuisine and label.de=Kueche. These two
labels signify how to display the hasCuisine field in the English and German
languages. The form field is a drop down menu of the Cuisine class and dis-
plays all the subclasses of Cuisine. Similarly, the hasName datatype property
is labeled Name and its field is a text box. Figure 4 illustrates this concept.
LATEX style file for Lecture Notes in Computer Science – documentation9
Fig.4. Ontology Form
4.2 Cascading Forms for Combined Service Queries
We have built a front-end that supports combined querying of services. In or-
der to accomplish this, the user interface needs to be able to display more than
one service form at a time. The front-end server downloads a number of ontolo-
gies for each service class. In order to allow multiple services to be searched
for in a single query, the relationship between these service classes needs to
be established. This is accomplished by passing the ontologies through an on-
tology mapper which establishes the relationship between the ontologies. We
implement a simple ontology mapping tool by storing the relationship between
services in a relationship table. This also includes the corresponding matching
users click on these links, cascading forms are displayed. The matching prop-
erty or properties are only displayed once and inserted as a shared property in
the combined query. Figure 5 below shows a combination of Restaurant and
10LATEX style file for Lecture Notes in Computer Science – documentation
Theater service forms displayed with one Neighborhood field which is the com-
Fig.5. Cascading Service Forms
4.3Creating GloServ Queries
Single service ontology queries in GloServ are a conjunction of all the prop-
erty values. Although more complicated queries can be formed (ie. disjunctive
queries, cardinality assignments, and set equivalences), a conjunctive query suf-
fices for the purpose of the demonstrative use cases. Besides creating simple
ontology queries, the front-end also checks to see if there are combined queries
or keywords entered and creates the query accordingly. The information en-
tered in by the user is converted to a GloServ query. For example, hasCuisine
some Italian, signifies that the object property hasCuisine is assigned to
For combined queries, the front-end creates a primary query out of the first
service class and nested queries from subsequent service classes. In the example
above, the Restaurant service query will be the primary one and the Theater
LATEX style file for Lecture Notes in Computer Science – documentation11
query, the nested one. If there were more services cascaded in the form then
these would be nested within each other.
For the collaborative user search, a single query is formed given inputs from
multiple users. The interface is synchronized across user group. Every time a
user enters a value for a property, her collaborators see an updated form with
the property value the user entered. Once every user inputs a value for a given
property, the query that is constructed is a conjunction of each property value
entered by the users.
The query is put in an XML message which indicates what type of query
it is (service registration or query). It is then passed onto the front-end server
which constructs the appropriate GloServ message and issues the query to the
4.4 Displaying Results
For single service query results, the results are displayed in a list where each
instance and its properties are displayed. In combined queries, results are dis-
played using an instance tree that shows the relationship between each service
instance. An example of this is the combined query of Restaurant and Theater,
where the tree is formed such that instances of the Restaurant class are parents
of matching Theater.
For location-based services, the results are also displayed in a Google map.
Each service instance is labeled with a pin on the Google map. For combined
service queries, the map shows the primary query tagged in red and the nested
query tagged in green. When clicking on either the red or green tags, the user
can choose to show only corresponding services. This grays out all the services
except for those services which match the current one that has been clicked. In
this way, users can easily navigate through the map and can determine what the
matching services are. Figure 6 shows how the query results are displayed for a
combined restaurant and movie search.
The front-end implementation is done mostly in PHP. For the collaborative in-
a Smarty template engine to generate HTML code for PHP . Smarty facili-
tates a manageable way to separate application logic and content from its pre-
sentation. The collaborative search interface requires user view synchronization
framework is used to communicate with the client side and web server. AJAX
provides a framework for creating efficient and interactive web applications.
12LATEX style file for Lecture Notes in Computer Science – documentation
Fig.6. Google map results for combined service queries
GloServ nodes communicate with each other using a form of property-
value-pair encoding. The implementation separates the message format and en-
coding from the logic. Figure 7 below shows how the messages are exchanged
the GloServ back-end interface. It is assumed that the front-end server already
has the ontology file in the cache.
6 Related Work
There are a few service discovery protocols in use today. Most service discovery
mechanisms are localized and use attribute-value pairs for service descriptions.
Below we describe each of these and compare them to GloServ.
LATEX style file for Lecture Notes in Computer Science – documentation 13
Fig.7. Query Code Walk
SLP  and Jini  are both similar in that they have agents that manage
services, users and directories of services. Agents advertise each others’ pres-
ence to each other using either multicast or unicast. In SLP, service registration
and queries are broadcast to the directory agents or directly between the ser-
vice and user agents depending on if the directory agents are present. In Jini,
however, a client downloads the service proxy and invokes through Java RMI
in order to access the service through a discovery process. Service descriptions
in SLP are done in simple attribute-value pairs whereas Jini matches interfaces.
SLP is mainly used in local area networks. Jini can span to a larger enterprise
networks. The Universal Description, Discovery and Integration (UDDI) 
specification is used to build discovery services on the Internet. UDDI provides
Services are described in XML and published using a Publisher’s API. Con-
sumers access services by using the Programmer’s API built on top of SOAP.
Services in UDDI are stored in a centralized business registry. The main draw-
back of UDDI is that it has a centralized architecture and does not span to a
14LATEX style file for Lecture Notes in Computer Science – documentation
Since these architectures are either centralized or scale to a local level,
GloServ provides the best architecture for service composition in a globally
distributed network. Also, because of its use of ontologies, a friendlier user in-
terface can be implemented which allows intelligent querying of services.
GloServ provides ontology service descriptions as well as a framework for dif-
ferent service classes to be aggregated in a single network. Due to these at-
tributes, we have built a web-based front-end to demonstrate the interesting use
cases for GloServ. These include searching for a combination of location-based
services, tagged services as well as collaborating with other users on a single
1. Asp.net ajax. http://www.asp.net/ajax/.
2. del.icio.us. http://del.icio.us/.
3. Flickr. http://www.flickr.com.
5. Owl: Web ontology language. http://www.w3.org/2004/OWL/.
6. Php hypertext processor. http://www.php.net/.
7. Smarty template engine. http://www.smarty.net.
8. Web services architecture. http://www.w3.org/TR/ws-arch/.
9. Zagat survey. http://www.zagat.com/.
10. Uddi technical white paper. white paper, uddi (universal description, discovery and integra-
tion). Technical report, OASIS, September 2000.
11. K. Arabshian, C. Dickmann, and H. Schulzrinne. Service composition in an ontology-based
global service discovery system. Technical report, Columbia University, New York, NY,
12. K. Arabshian and H. Schulzrinne. Combining ontology queries with key word search in ser-
vice discovery. In ACM/IFIP/USENIX 8th International Middleware Conference, Newport
Beach, California, 2007.
13. K. Arabshian and H. Schulzrinne. An ontology-based hierarchical peer-to-peer global ser-
vice discovery system. Journal of Ubiquitous Computing and Intelligence (JUCI), 2:133–
14. U. D. DI. UDDI technical white paper. White paper, UDDI (Universal Description, Discov-
ery and Integration), September 2000.
15. E. Guttman, C. Perkins, J. Veizades, and M. Day. Service location protocol, version 2. RFC
2608, Internet Engineering Task Force, June 1999.
16. Sun Microsystems. Jini architectural overview. Technical report, 1999.
17. S. Ratnasamy, P. Francis, M. Handley, R. Karp, and S. Shenker. A scalable content address-
able network. In Proceedings of ACM SIGCOMM 2001, San Diego, CA, 2001.