=Paper=
{{Paper
|id=Vol-540/paper-4
|storemode=property
|title=Rethinking the Semantic Annotation of Services
|pdfUrl=https://ceur-ws.org/Vol-540/ugs2009_submission_4.pdf
|volume=Vol-540
}}
==Rethinking the Semantic Annotation of Services==
https://ceur-ws.org/Vol-540/ugs2009_submission_4.pdf
Rethinking the Semantic Annotation of Services
Nikolaos Loutas1,2, Vassilios Peristeras1 and Konstantinos Tarabanis2
1
National University of Ireland, Galway, Digital Enterprise Research Institute
{firstname.lastname}@deri.org
2
Information Systems Lab, University of Macedonia, Thessaloniki, Greece
{nlout, kat}@uom.gr
Abstract. This work extends and enhances existing semantic service models by
involving users and by including service metadata related to the user’s view of
the service and their behaviour. We borrow ideas and extend the models and
practices for the annotation of Web content and information resources that has
recently become popular in widely-used social platforms. Users are encouraged
to describe in their own terms the services they use. Our approach strengthens
user participation in the Web and more generally in the service industry by
providing service metadata, which are later used as a form of lightweight user-
side semantic annotation of services. This annotation is provided explicitly by
the users and/or implicitly by identifying patterns in the users' behaviour. This
type of service annotation acts supplementary to the service descriptions
provided by the service providers and is linked to the actual use of the services.
Finally, we harvest the collected metadata and use it for facilitating discovery
and clustering of services, as well as to enable service recommendations and
matchmaking with users’ profiles.
Keywords: service, SOA, semantic, social contract, social metadata
1 Introduction
The Service Oriented Architecture (SOA) paradigm is the prevalent way of
developing enterprise Information Systems (IS). The main idea behind SOA is the
ability to (re)use and share services from different sources. The introduction of Web
Services (WS) contributed significantly to SOA’s commercial uptake and helped
SOA adoption by industry. Further on IBM proposed a reference architecture for
SOA which identifies three basic entities: the service provider, the service requestor
and the service broker. This reference architecture supports four fundamental
functionalities, namely service discovery, composition, publishing and invocation.
There has been a lot of debate with respect to the advantages and disadvantages of
SOA. The advantages of SOA include enhanced organizational agility and
architectural flexibility, reduced costs and increased Return on Investment [18, 23].
Nevertheless, the real added value of SOA implementations is still questioned as
technological, organizational and financial barriers hinder the uptake of SOA [3].
Semantics have been applied in SOA as a means to enhance the WS brokerage
model. The idea is that semantically described WS could enable and facilitate the
dynamic discovery, invocation, execution, composition and monitoring of services at
run-time [10]. This led to the definition of various service ontologies and Semantic
Web Services (SWS) models starting from DAML-S, continuing to OWL-S [11],
WSMO [17], WSDL-S [13] and followed by SAWSDL [5].
In this paper, we add an additional layer to existing semantic approaches. Apart
from the formal service descriptions (semantic or not) that are made available by the
service providers in SOA environments, we enhance the service descriptions by
“capturing” and “attaching” information, which is related to the actual usage of the
services by users in the real world. Our intention is to add a social layer on top of
SOA. We propose two main mechanisms for this:
• We allow users to annotate the services they use.
• We analyze the behavioural/service usage patterns of the users’, i.e. how the
users consume the services, by monitoring their behaviour and monitoring
their actions.
Afterwards, the information that we collect is used to enrich the existing service
descriptions. We call this process social annotation of services and the metadata that
emerge bottom-up from this process social metadata. The social annotation may be
supplementary to the semantic service descriptions that are already provided by the
service providers.
In order to accommodate the social aspects and characteristics of services, existing
semantic service models need to be extended. In this paper, we discuss the required
extensions; we introduce an approach for adding social annotations to services and
demonstrate how to utilize the extra (social) semantics that emerge in order to
facilitate service search, mashing and recommendation.
Before continuing, it is necessary to clarify that our approach focuses mainly on
extrovert services which have a business value and are to be used by the end-users.
The remainder of this paper is organized as follows: section 2 presents our
motivation. Section 3 discusses in detail the idea of social annotation of services and
introduces the notion of the Social Contract. Section 4 shows how an existing
semantic service models can be extended in order to include social metadata. Section
5 describes our prototype. Finally, section 6 concludes the paper and discusses our
research directions.
2. Motivation
All the semantic service frameworks proposed so far, e.g. WSMO, SAWSDL, OWL-
S, both complex and lightweight, share a common principle: they assume that the
(semantic) description of services comes solely from the service providers. This
results into two serious limitations:
• The users are totally left outside of the service description process. These
approaches do not take into account the way that the users of the service
perceive it. For example, users cannot detail the reason they use a service,
e.g. to book a flight, or the context in which the service is used, e.g. a service
as part of a more general “travelling” context. Recently, both the research
community and industry sensed this shortcoming. There are already attempts
which focus on the user’s perspective which was left completely out of
picture in SOA e.g. [12].
• Semantic service efforts have still a low adoption rate. To come up with
elaborated semantic service descriptions, the service providers have to be
convinced about the additional value of these semantics in order to spend
resources to annotate their services. However at the moment, it seems rather
unlikely to convince service providers to use existing SWS frameworks, and
as a result these efforts are not taking off.
So far, there has been no real large-scale application of SWS in industry. Among
other problems in [24], the author claims that SWS have ill-defined semantics and
that service ontologies usually describe the semantics of WSDL interfaces, which are
different from the semantics of the WS. As such the existing SWS approaches have
not managed to fully support automated discovery, matchmaking, composition, and
execution. Generally, the high complexity of the SWS approaches discourages both
technical and business people from adopting such solutions. These problems have
created pressure to the SWS research community to come up with lightweight
approaches, which may lack in expressivity but win in simplicity (e.g. SAWSDL, SA-
REST [20], WSMO-Lite [22]). However, it is still early to evaluate the applicability
and the adoption of such light semantic service models.
Lately, Web 2.0 is emerging as a new computing paradigm. Web 2.0 preaches for
active user participation in the Web through user-centric Web portals and applications
[14]. In Web 2.0 there is no clear line of separation between service providers and
users as the latter interact with the Web not just as information receivers but also as
content providers [4]. In Web 2.0 platforms users add data and metadata: they add
content, e.g. photos, multimedia and documents, and then use tags for attaching
meaning to this content. Other users also add metadata (tags), which are then used to
enable better search (e.g. [2, 25]) and discovery, personalization of the user’s
experience etc.
Moreover, unlike SOA environments, in Web 2.0 semantics (metadata) come
mainly from tags and folksonomies and as such emerge in a bottom-up fashion
directly from the users (e.g. [7, 21]). We argue that as this user-defined metadata are
used for creating richer descriptions for resources (e.g. photos, files etc.), they could
be likewise applied to services for enhancing their descriptions. This metaphor is
challenging and capitalizes on the view that Web 2.0 and SOA are two converging
and complementary paradigms [19].
Within Web 2.0, new types of services appear which do not follow the SOA
principles and are created in a decentralized manner. These services are usually
generated by users and not by service engineers. For example, mashups, are
introduced as a new simple way of composing services and combining content from
different sources. Interestingly enough, Web 2.0 services lack a standardized
description from their providers, as there are neither standardized ways to describe
services nor public repositories to store these descriptions. The services are generated
in a completely decentralized and uncontrolled way and the overall architecture lacks
the core SOA idea of a WS broker that mediates between the service providers and
service clients.
Thousands of mashups are available at the moment and all indications show that
this number may soon scale to millions. All these mashups are capable to provide
access to huge amount of distributed content and/or services. But how one can find
the mashup (s)he really needs, when (s)he needs it? How can a driver that is running
out of gas find if there is a mashup that displays on a map the open gas stations in the
area where (s)he is driving? How can the four basic SOA functionalities, namely
discovery, composition, publishing and invocation be achieved in a Web 2.0
environment? Currently the means, e.g. models, architectures and applications for
answering such questions are not available.
Thus, in this work we will show how social metadata can be included in service
descriptions, let them be SOA services or RESTful services, in order to facilitate
service search, mashing and recommendation.
3. Social Annotation of Services
A service can be seen from different perspectives starting from a complex business
process and going down to software component [1]. In [22] five complementary parts
of a service’s description, called service contracts, are defined, namely:
• The Information Model which refers to the data model that is used to
semantically describe the service inputs, outputs and fault messages.
• The Functional Descriptions which describe the service’s functionalities.
• The Non-Functional Descriptions which define details related to the
implementation or the running environment of the service, e.g. name, author,
URL, version.
• The Behavioral Descriptions which define the service’s choreography and
internal workflow.
• The Technical Descriptions which define details regarding the format of the
messages, the communication protocols and the access points of the service.
In our previous work, we reviewed several SWS efforts. We observed that they can
be grouped into two distinct categories:
• SWS frameworks which refer to formal and complex efforts for semantically
annotating services, and
• Semantic service models which refer to emerging lightweight approaches for
semantically annotating services.
In the first category we find OWL-S, SWSF, WSMO, IRS-III and WSMO-Lite,
while WSDL-S and its successor SAWSDL, SA-REST and MicroWSMO are placed
under the second category. Afterwards, we examined how the different SWS efforts
address the service contracts described above.
Summarizing our comparative analysis, we found out that SWS frameworks
provide the language for encoding the ontologies that form their Information Model.
On the contrary, semantic service models allow the use of any ontological language.
Thus, their Information Model may be comprised by a set of ontologies encoded
using different languages. Moreover, both the Functional and the Behavioural
Descriptions can be expressed in detail using a SWS framework. Although the
semantic service models provide some means for specifying services’ functionalities,
these mechanisms lack in expressivity. Finally, regarding the Technical Descriptions
both SWS frameworks and semantic service models rely mainly on the WSDL
specification, excluding the case of SA-REST that refers to RESTful services, thus no
WSDL file exists.
In all aforementioned approaches, the metadata of all the service contracts come
solely from the service providers. In this work we argue that it will be highly
beneficial both for users and for service providers, if the semantic descriptions of the
services were enriched with information that comes from the actual usage of a service
in a bottom-up fashion. The semantic description in this case emerges from the usage
of the service, thus giving a social aspect to service annotation.
To achieve this, we introduced a new service contract, which we call social
contract [9]. The social contract captures the way that the users’ perceive the service,
when and why they use the service etc. The social contract has been until now
neglected when modelling and implementing services. The social metadata of services
may derive explicitly or implicitly:
• Explicitly, where metadata is added by users who wish to describe the service
in their own terms. Users can annotate services, similarly to what they
currently do for products, content and multimedia in platforms like Flickr or
YouTube. Users would describe why they use a service, for what reason, on
which occasion etc. They could also add annotations that are related with
attributes of the service like inputs or outputs. In other cases, users may
express their satisfaction or dissatisfaction with regards to quality, usability,
user-friendliness etc. However, some of the service attributes still remain to
be solely described by the service provider, e.g. the Technical Descriptions.
We propose the use of tags and tagging mechanisms as the means for
enabling explicit service annotation. Tagging is easy and straightforward and
users are already familiar with it. Tagging mechanisms can either rely
completely on free text or it can be further supported by vocabularies and
ontologies. Here lies a substantial difference between the annotations from
the service providers and those that come from users. Service providers are
more likely to use formal service models combined with structured
vocabularies and ontologies, while users would rather use tags.
• Implicitly, where information about the service can be inferred by monitoring
the user’s behaviour while using the service and then enrich the service
description. For example, imagine that a statistically significant number of
users execute service B after service A or users with similar profiles have an
interest for some particular services. In both cases, this information can
provide us with interesting usage patterns that can be further exploited e.g.
for a service recommendation system.
4. Including social metadata in existing semantic service models
Until know we have discussed at a conceptual level how social metadata could be
included in existing semantic service models. In this section we will show how we
applied this idea by extending an existing semantic service model, namely SA-REST.
SA-REST [20] introduces a lightweight approach for adding semantics to RESTful
WS. It assumes that it is highly likely that when a WS is made available online, the
provider will release an (X)HTML description of the service as well. Thus, SA-REST
suggests to semantically annotate this (X)HTML service description using RDFa or
GRDDL. SA-REST uses a set of predefined elements for annotating different
attributes of a service, e.g. input, output and sem-class.
SA-REST is simple and easily extendable, mainly due to the fact that it is based on
RDFa. Moreover, the lightweight semantic annotation of services, as suggested by
SA-REST, fits very well with our line of work. Moreover, as the number of RESTful
services is increasing influenced by the growth of Web 2.0, we expect lightweight
efforts to semantically annotate services using RDFa to become popular. Thus, we
chose SA-REST in order to semantically annotate the services of our portal and
capture and include social metadata.
SA-REST like other similar approaches for semantically annotating services covers
the five service contracts but does not support the social contract. This means that SA-
REST assumes that the user has a passive role and that the service descriptions come
only from the service provider. Summarizing, in SA-REST the information model of
a service is identified by its inputs and outputs. Both the behavioural and the
functional Descriptions are weakly defined in the SA-REST model due to the fact that
SA-REST aims at providing a lightweight approach for service descriptions
sacrificing its expressivity. Finally, technical information can be represented by the
method and protocol elements, while, non-functional descriptions can be derived from
the domain-rel, sem-class and sem-rel elements.
In order to include social metadata in SA-REST, we first express social metadata
by means of a simple folksonomy and then make use of the sem-class element in
order to create a link between a service and this folksonomy. In order to indicate that
this sem-class element refers to the social metadata of this service, we make use of
the typeof element and make a reference to the social contact concept of our model.
It is worth mentioning that the social contract, as introduced in this work, can be
combined with any other SWS framework. For example, WSMO non-functional
properties Web Services or a SAWSDL model references enable the inclusion of
social metadata in WSMO and SAWSDL services respectively.
5. Prototype
For the needs of our prototype, we have scoped our focus on the eGovernment
domain. We have selected eGovernment as our application domain due to previous
work and expertise in the domain, but also because eGovernment is a challenging
test-bed with thousands of services provided worldwide by public agencies to billions
of clients. Another interesting characteristic of public administration is that its
clientele is not restricted to a certain group of people with common needs or interests.
In fact, public administration tries to cover the needs of practically each and every
citizen. To achieve this, public administration tries to group the diverse needs of its
clientele and translate them into services.
Social annotation of services can provide valuable input for public administration
during the public service design process that could lead to higher quality services that
would fit better to the clients’ needs. Furthermore, public administration could predict
future needs of its clientele, based on trends expressed through the social annotation
of services. This will allow public administration to be more agile and proactive.
Our eGovernment portal plays the role of a national entry point to the services
provided by public administration. In fact a prototype of the portal, which is available
at http://195.251.218.39/cyprus, is currently used in a pilot study in Cyprus.
Citizens use this portal to get information about public services. The descriptions of
the services that are made available via the portal are semantically annotated using the
extended semantic service model that is presented in section 4. Apart from the
inclusion of social metadata, we have also included eGovernment domain specific
semantics in the semantic descriptions of our services. Towards this direction, the
GEA Public Service model [15, 16] was employed, which introduces a conceptual
representation of a public service. As such, it introduces core concepts of a public
service, such as service input and output, service provider, service preconditions,
service domain etc. We used RDFain order to include in our semantic eGovernment
service descriptions eGovernment domain specific semantics (Table 1). The semantic
descriptions are initially automatically created on the service provider’s side, but once
they are released and used by the users, they are enriched and become more
expressive as social metadata are added.
In our social eGovernment portal users can tag the services provided or use the tags
already given by other users. Moreover, the users’ behaviour while using the portal,
e.g. when navigating from one service to another, is anonymously monitored. This
allows us to extract social metadata from their behavioural patterns without violating
their privacy.
In both cases mentioned above, social metadata is extracted, which is then used for
enhancing the user’s experience in the portal. Users can browse the underlying
eGovernment service repository using different mechanisms, e.g. tags/tagclouds,
keyword search. Moreover, they can get recommendations about services that are
popular or beneficial or fit their profiles or are related with services that they have
already used. A detailed description of the portal can be found in [8].
Table 1. Example of service description annotated with the extended SA-REST model
The Driving License Issuance public service is provided
by the
Regional Authorities and belongs to the
Community and Social Services domain
6. Conclusions and Future Work
In this paper we discussed the need for extending the existing SOA-based service
descriptions by including socially derived metadata. We add a “social” layer on top of
existing semantic or non-semantic SOAs. This layer enriches the service description
and facilitates the discovery and automatic clustering of services.
Two mechanisms for capturing this metadata have been proposed, receiving the
annotations explicitly and directly by the users through tagging and implicitly through
analysis of users’ behaviour and usage patterns. In order to validate our arguments an
extension of SA-REST which accommodates social metadata was proposed. In
addition to that a social eGovernment portal was developed.
Harvesting the knowledge that can be extracted from social the social metadata of
services is expected to benefit both users and service providers. From the user’s
perspective, it allows them to:
• Express their view of the services they use by annotating these services, e.g.
using tags.
• Personalize their service searches, thus improving the quality and the
coherence of the result set and bridging the service discovery gap [6]. Social
metadata, e.g. tags, can be added as criteria to search queries, thus narrowing
down the result sets.
• Get recommendations about related services. Service platforms may suggest
to their users services that share common tags.
• Form communities of interest and/or practice. As described earlier, social
semantics can also be extracted by monitoring the users’ behaviour. In this
case, users that tend to use similar services can be grouped in communities
of interest. For example, a community of users who use online collaboration
services.
Service providers can exploit the social annotation of services to:
• Improve the classification and clustering of services according to their
functionality/output/behaviour etc. Since the semantics that emerge from the
social annotation of services stem directly from the users, they can create a
bottom-up classification of services.
• Get feedback on their services and improve their quality or design new
services in order to cover emerging customers’ needs.
The benefits that social metadata are anticipated to have both to uses and service
providers, set also the pillars of our future research plan. Hence, we will try draft and
develop possible applications and prototypes and enhance existing systems by
including and harvesting social metadata.
Acknowledgments
This work is supported by Science Foundation Ireland under grant SFI/08/CE/I1380 (Líon 2).
References
1. Baida, Z., Gordijn, J., Omelayenko, B.: A shared service terminology for online service
provisioning. In: ICEC 2004, pp. 1-10 (2004)
2. Bao, S., Xue, G., Wu, X., Fei, B., Su, Z.: Optimizing web search using social annotations.
In: 16th World Wide Web Conference, pp. 501-510 (2007)
3. Bloomberg, J.: Why is SOA taking so long?, ZapThink.
http://searchsoa.techtarget.com/tip/0,289483,sid26_gci1132084,00.html (2005)
4. Domingue, J., Fensel D., González-Cabero, R.: SOA4All, Enabling the SOA Revolution
on a World Wide Scale. In: 2nd IEEE International Conference on Semantic Computing
(2008)
5. Farrell, J., Lausen, H.: Semantic Annotations for WSDL and XML Schema”, W3C
Recommendation, http://www.w3.org/TR/sawsdl/ (2007)
6. Fernandez, A., Hayes, C., Loutas, N., Peristeras, V., Polleres, A., Tarabanis, K.: Closing
the Service Discovery Gap by Collaborative Tagging and Clustering Techniques. In:
International Semantic Web Conference, Workshop on Service Discovery and Resource
Retrieval in the Semantic Web (2008)
7. Gruber, T.: Ontology of Folksonomy: A Mash-Up of Apples and Oranges. International
Journal of Semantic Web and Information Systems 3, 1–11 (2007)
8. Kopouki Papathanasiou, M., Loutas N., Peristeras V., Tarabanis K., Combining service
models, semantic and Web 2.0 technologies to create a rich citizen experience. In: 2nd
World Summit on the Knowledge Society, WSKS (1): 296-305 (2009)
9. Loutas, N., Peristeras, V., Tarabanis, K.: Extending Service Models to Include Social
Metadata. In: WebSci'09: Society On-Line (2009)
10. Martin, D., Domingue, J.: Semantic Web Services, Part 2. IEEE Intelligent Systems 22(6),
8-15 (2007)
11. Martin, D., Burstein, M., Hobbs, J., Lassila, O., McDermott, D., McIlraith, S., Narayanan,
S., Paolucci, M., Parsia, B., Payne, T., Sirin, E., Srinivasan, N., Sycara, K.: OWL-S:
Semantic Markup for Web Services, W3C Member Submission,
http://www.w3.org/Submission/OWL-S/ (2004)
12. Meyer, H., Weske, M.: Light-Weight Semantic Service Annotations through Tagging. In:
Service-Oriented Computing 4294, pp. 465-470 (2006)
13. Miller, J., Verma, K., Rajasekaran, P., Sheth, A., Aggarwal, R., Sivashanmugam, K.:
WSDL-S: Adding Semantics to WSDL - White Paper. Large Scale Distributed
Information Systems, http://lsdis.cs.uga.edu/library/download/wsdl-s.pdf (2004)
14. Murugesan, S.: Understanding Web 2.0. IT Professional 9(4), 34-41 (2007)
15. Peristeras, V., Tarabanis, K.: The Governance Architecture Framework and Models. In:
Saha, P. (ed.) Advances in Government Enterprise Architecture. IGI (2008)
16. Peristeras, K. Tarabanis, K.: Towards an Enterprise Architecture for Public
Administration: A Top Down Approach. European Journal of Information Systems 9, 252-
260 (2002)
17. Roman, D. Keller, U., Lausen, H., Bruijn, J.d., Lara, R., Stollberg, M., Polleres, A., Feier,
C., Bussler, C., Fensel, D.: Web Service Modeling Ontology, Applied Ontology 1(1), 77-
106 (2005)
18. Schmelzer, R.: The ROI of SOA. ZapThink,
http://www.zapthink.com/report.html?id=ZAPFLASH-20050127 (2005)
19. Schroth, C., Janner, T.: Web 2.0 and SOA: Converging Concepts Enabling the Internet of
Services. IEEE IT Professional 9(3), 36-41 (2007)
20. Sheth, A., Gomadam, K.. Lathem, J.: SA-REST: Semantically Interoperable and Easier-to-
Use Services and Mashups. IEEE Internet Computing 11(6), 91-94 (2007)
21. Specia, L., Motta, E.: Integrating Folksonomies with the Semantic Web. In: European
Semantic Web Conference 2007, pp. 624-639 (2007)
22. Vitvar, T., Kopecky, J., Fensel, D.: WSMO-Lite: Lightweight Semantic Descriptions for
Services on the Web. CMS WG Working Draft (2008)
23. Windley, P.: Teaming up for SOA. Infoworld, pp 25-26 (2007)
24. Xuan S.: Semantic Web Services: An Unfulfilled Promise. IEEE IT Professional 9(4), 42-
45 (2007)
25. You, G., Hwang, S.: Search structures and algorithms for personalized ranking. Journal of
Information Sciences 178(20), 3925-3942 (2007)