A Survey of Automated Web Service Composition Methods

Abstract

In today's Web, Web services are created and updated on the fly. It's already beyond the human ability to analysis them and gen- erate the composition plan manually. A number of approaches have been proposed to tackle that problem. Most of them are inspired by the re- searches in cross-enterprise workflow and AI planning. This paper gives an overview of recent research efforts of automatic Web service compo- sition both from the workflow and AI planning research community.

Full text

A Survey of Automated Web Service
Composition Methods
Jinghai Rao and Xiaomeng Su
Norwegian University of Science and Technology,
Department of Computer and Information Science,
N-7491, Trondheim, Norway
{jinghai, xiaomeng}@idi.ntnu.no
Abstract. In today’s Web, Web services are created and updated on
the fly. It’s already beyond the human ability to analysis them and gen-
erate the composition plan manually. A number of approaches have been
proposed to tackle that problem. Most of them are inspired by the re-
searches in cross-enterprise workflow and AI planning. This paper gives
an overview of recent research efforts of automatic Web service compo-
sition both from the workflow and AI planning research community.
1 Introduction
Web services are considered as self-contained, self-describing, modular applica-
tions that can be published, located, and invoked across the Web. Nowadays,
an increasing amount of companies and organizations only implement their core
business and outsource other application services over Internet. Thus, the ability
to efficiently and effectively select and integrate inter-organizational and hetero-
geneous services on the Web at runtime is an important step towards the devel-
opment of the Web service applications. In particular, if no single Web service
can satisfy the functionality required by the user, there should be a possibility
to combine existing services together in order to fulfill the request. This trend
has triggered a considerable number of research efforts on the composition of
Web services both in academia and in industry.
In the research related to Web services, several initiatives have been con-
ducted with the intention to provide platforms and languages that will allow easy
integration of heterogeneous systems. In particular, such languages as Universal
Description, Discovery, and Integration (UDDI) [4], Web Services Description
Language (WSDL) [9], Simple Object Access Protocol (SOAP) [6] and part of
DAML-S [14] ontology (ServiceProfile and ServiceGrounding), define standard
ways for service discovery, description and invocation (message passing). Some
other initiatives such as Business Process Execution Language for Web Service
(BPEL4WS) [2] and DAML-S ServiceModel, are focused on representing service
compositions where flow of a process and bindings between services are known
a priori.
J. Cardoso and A. Sheth (Eds.): SWSWPC 2004, LNCS 3387, pp. 43–54, 2005.
c
Springer-Verlag Berlin Heidelberg 2005

44 J. Rao and X. Su
Despite all these efforts, the Web service composition still is a highly com-
plex task, and it is already beyond the human capability to deal with the whole
process manually. The complexity, in general, comes from the following sources.
First, the number of services available over the Web increases dramatically dur-
ing the recent years, and one can expect to have a huge Web service repository
to be searched. Second, Web services can be created and updated on the fly, thus
the composition system needs to detect the updating at runtime and the decision
should be made based on the up to date information. Third, Web services can
be developed by different organizations, which use different concept models to
describe the services, however, there does not exist a unique language to define
and evaluate the Web services in an identical means.
Therefore, building composite Web services with an automated or semi-
automated tool is critical. To that end, several methods for this purpose have
been proposed. In particular, most researches conducted fall in the realm of
workflow composition or AI planning.
For the former, one can argue that, in many ways, a composite service is
similar to a workflow [8]. The definition of a composite service includes a set of
atomic services together with the control and data flow among the services. Sim-
ilarly, a workflow has to specify the flow of work items. The current achievements
on flexible workflow, automatic process adaption and cross-enterprise integration
provide the means for automated Web services composition as well. In addition,
the dynamic workflow methods provide the means to bind the abstract nodes
with the concrete resources or services automatically.
On the other hand, dynamic composition methods are required to generate
the plan automatically. Most methods in such category are related to AI plan-
ning and deductive theorem proving. The general assumption of such kind of
methods is that each Web service can be specified by its preconditions and ef-
fects in the planning context. Firstly, a Web service is a software component
that takes the input data and produces the output data. Thus the preconditions
and effects are the input and the output parameters of the service respectively.
Secondly, the Web service also alters the states of the world after its execution.
So the world state pre-required for the service execution is the precondition, and
new states generated after the execution is the effect. A typical example is a
service for logging into a Web site. The input information is the username and
password, and the output is a confirmation message. After the execution, the
world state changes from “not logged in” to “logged in”. The “logged in” state
will be keeping until the “log out” service is invoked. If the user can specify the
preconditions and effects required by the composite service, a plan or process
is generated automatically by logical theorem prover or AI planners without
knowledge of predefined workflow. During the planning, the business logic can
provide constraints in the planning setting.
In this paper we will present an overview of recent methods that provide
automation to Web service composition. The automation means that either the
method can generate the process model automatically, or the method can locate
the correct services if an abstract process model is given. Some methods based on

A Survey of Automated Web Service Composition Methods 45
workflow have been reported in the work by Benatallah [5], but to our knowledge,
no overview on service composition methods related to AI planning has been
published yet. As a result, in the paper, we will have more focus on the AI
planning methods than the workflow based methods.
This paper is organized as follows. Section 2 presents an abstract framework
for Web service composition. Section 3 is the introduction of automatic Web
service composition based on workflow methods. Section 4 provides an overview
and comparison for the selected composition methods based on AI planning. The
last section concludes the paper.
2 Web Services Composition Framework
Here, we propose a general framework for automatic Web services composition.
This framework is in high-level abstraction, without considering a particular
language, platform or algorithm used in composition process. The aim of the
framework is to give the basis to discuss similarities and differences of the avail-
able service composition methods. In addition, we also use the framework to
unify the terms used in the paper.
Fig. 1. The framework of the service composition system
A general framework of the service composition system is illustrated in Fig. 1.
The composition system has two kinds of participants, service provider and ser-
vice requester. The service providers propose Web services for use. The service
requesters consume information or services offered by service providers. The
system also contains the following components: translator, process generator,
evaluator, execution engine and service repository. The translator translates be-
tween the external languages used by the participants and the internal languages
used by the process generator. For each request, the process generator tries to
generate a plan that composes the available services in the service repository
to fulfill the request. If more than one plan is found, the evaluator evaluates all
plans and proposes the best one for execution. The execution engine executes
the plan and returns the result to the service provider.
Most precisely, the process of automatic service composition includes the
following phases:

46 J. Rao and X. Su
Presentation of Single Service: firstly, the service providers will advertise
their atomic services at a global market place. There are several languages avail-
able for advertising, for example, UDDI [4] or DAML-S ServiceProfile [14]. The
essential attributes to describe a Web service include the signature, states and
the non-functional values. The signature is represented by the service’s inputs,
outputs and exceptions. It provides information about the data transformation
during the execution of a Web service. The states are specified by precondition
and postcondition. We model it as the transformation from one set of states to
another in the world. Non-functionality values are those attributes that are used
for evaluating the services, such as the cost, service quality and security issues.
Translation of the Languages: most service composition systems distinguish
between the external and internal service specification languages. The external
languages are used by the service users to enhance accessibility of the users in
the sense that the users can express what they can offer or what they want in
a relatively easy manner. They are usually different from the internal ones that
are used by the composition process generator, because the process generator re-
quires more formal and precise languages, for example, the logical programming
languages. So far, the users have already get used to the standard Web service
languages, such as WSDL and DAML-S. Thus the translation components be-
tween the standard Web service languages and the internal languages have to
be developed.
Generation of Composition Process Model: in the meantime, the service
requester can also express the requirement in a service specification language. A
process generator then tries to solve the requirement by composing the atomic
services advertised by the service providers. The process generator usually takes
the functionalities of services as input, and outputs process model that describes
the composite service. The process model contains a set of selected atomic ser-
vices and the control flow and data flow among these services.
Evaluation of Composite Service: it is quite common that many services
have the same or similar functionalities. So it is possible that the planer generates
more than one composite service fulfilling the requirement. In that case, the
composite services are evaluated by their overall utilities using the information
provided from the non-functional attributes. The most commonly used method is
utility functions. The requester should specify weights to each non-functionality
attributes and the best composite service is the one who is ranked on top.
Execution of Composite Service: after a unique composite process is se-
lected, the composite service is ready to be executed. Execution of a composite
Web service can be thought as a sequence of message passing according to the
process model. The dataflow of the composite service is defined as the actions
that the output data of a former executed service transfers to the input of a
later executed atomic service.
In the following we will give a survey on the methods used for the process
generator to generate the process. The methods can be either fully automated
or semi-automated.

A Survey of Automated Web Service Composition Methods 47
3 Web Service Composition Using Workflow Technique
In the workflow-based composition methods, we should distinguish the static and
dynamic workflow generation. The static one means that the requester should
build an abstract process model before the composition planning starts. The
abstract process model includes a set of tasks and their data dependency. Each
task contains a query clause that is used to search the real atomic Web service to
fulfill the task. In that case, only the selection and binding of atomic Web service
is done automatically by program. The most commonly used static method is to
specify the process model in graph. On the other hand, the dynamic composi-
tion both creates process model and selects atomic services automatically. This
requires the requester to specify several constraints, including the dependency
of atomic, the user’s preference and so on.
EFlow[7] is a platform for the specification, enactment and management of
composite services. EFlow uses a static workflow generation method. A com-
posite service is modeled by a graph that defines the order of execution among
the nodes in the process. The graph is created manually but it can be updated
dynamically. The graph may include service, decision and event nodes. Service
nodes represent the invocation of an atomic or composite service, decision nodes
specify the alternatives and rules controlling the execution flow, and event nodes
enable service processes to send and receive several types of events. Arcs in the
graph denote the execution dependency among the nodes. Although the graph
should be specified manually, EFlow provides the automation to bind the nodes
with concrete services. The definition of a service node contains a search recipe
that can be used to query actual service either at process instantiation time or at
runtime. As the service node is started, the search recipe is executed, returning
a reference to a specific service. In particular, the search recipe is resolved each
time when a service node is activated. They do so because the availability of ser-
vices may change very frequently in a highly dynamic environment. In [8], the
authors further refine the service composition platform and propose a prototype
of composite service definition language(CSDL). An interesting feature of CSDL
is that it distinguishes between invocation of services and operations within a
service. It provides the adaptive and dynamic features to cope with the rapidly
evolving business and IT environment in which Web services are executed.
Polymorphic Process Model (PPM)[23] uses a method that combines the
static and dynamic service composition. The static setting is supported by ref-
erence process-based multi-enterprise processes, the processes that consist of
abstract subprocesses, i.e., subprocesses that have functionality description but
lack implementation. The abstract subprocesses are implemented by service and
bined at runtime. This is similar to the service binding in EFlow. The dynamic
part of PPM is supported by service-based processes. Here, a service is modeled
by a state machine that specifies that possible states of a service and their tran-
sitions. Transitions are caused by service operation(also called service activity)
invocations or internal service transitions. In the setting, the dynamic service
composition is enabled by the reasoning based on state machine.

48 J. Rao and X. Su
4 Web Service Composition Using AI Planning
Many research efforts tackling Web service composition problem via AI planning
have been reported. In general, a planning problem can be described as a five-
tuple S, S0,G,A,Γ, where Sis the set of all possible states of the world, S0⊂S
denotes the initial state of the world, G⊂Sdenotes the goal state of the world
the planning system attempts to reach, Ais the set of actions the planner can
perform in attempting to change one state to another state in the world, and
the translation relation Γ⊆S×A×Sdefines the precondition and effects for
the execution of each action.
In the terms of Web services, S0and Gare the initial states and the goal states
specified in the requirement of Web service requesters. Ais a set of available
services. Γfurther denotes the state change function of each service.
DAML-S (also called OWL-S in the most recent versions) is the only Web
service language that announces the directly connection with AI planning. The
state change produced by the execution of the service is specified through the
precondition and effect properties of the ServiceProfile in DAML-S. Precondition
presents logical conditions that should be satisfied prior to the service being
requested. Effects are the result of the successful execution of a service. Since
DAML+OIL, the language used to build DAML-S, uses Description Logics [10]
as its logical foundation, DAML+OIL has the express power allowing for logical
expressions. The majority of the methods reported in this survey use DAML-S
as the external Web service description language. There are also a couple of
methods that use WSDL or their own languages.
In the following we introduces a list of Web service composition methods
based on AI planning. This kind of methods have been reported frequently in
recent years, so we can not claim that we have an exhaustive list of the methods.
We further classify the methods into five categories, namely, the situation cal-
culus, the Planning Domain Definition Language (PDDL), rule-based planning,
the theorem proving and others.
4.1 Situation Calculus
McIlraith et. al. [17, 19, 16] adapt and extend the Golog language for automatic
construction of Web services. Golog is a logic programming language built on
top of the situation calculus. The authors address the Web service composition
problem through the provision of high-level generic procedures and customiz-
ing constraints. Golog is adopted as a natural formalism for representing and
reasoning about this problem.
The general idea of this method is that software agents could reason about
Web services to perform automatic Web service discovery, execution, composi-
tion and inter-operation. The user’s request (generic procedure) and constraints
can be presented by the first-order language of the situation calculus(a logical
language for reasoning about action and change). The authors conceive each
Web service as an action - either a PrimitiveAction or a ComplexAction. Primi-
tive actions are conceived as either world-altering actions that change the state

A Survey of Automated Web Service Composition Methods 49
of the world or information-gathering actions that change the agent’s state of
knowledge. Complex actions are compositions of individual actions. The agent
knowledge base provides a logical encoding of the preconditions and effects of
the Web service actions in the language of the situation calculus. The agents use
procedural programming language constructs composed with concepts defined
for the services and constraints using deductive machinery. A composite service
is a set of atomic services which connected by procedural programming language
constructs(if-then-else, while and so forth).
The authors also propose a way to customize Golog programs by incorpo-
rating the service requester’s constraints. For example, the service requester can
use the nondeterministic choice to present which action is selected in a given
situation, or use the sequence construct to enforce the execution order between
two action. The generation of the plan have to obey the predefined constraint.
4.2 PDDL
A strong interest to Web service composition from AI planning community could
be explained roughly by similarity between DAML-S and PDDL representations.
PDDL is widely recognized as a standardized input for state-of-the-art plan-
ners. Moreover, since DAML-S has been strongly influenced by PDDL language,
mapping from one representation to another is straightforward (as long as only
declarative information is considered). When planning for service composition
is needed, DAML-S descriptions could be translated to PDDL format. Then
different planners could be exploited for further service synthesis.
In presenting the Web service composition method based on PDDL, McDer-
mott [15] introduces a new type of knowledge, called value of an action, which
persists and which is not treated as a truth literal. From Web service construction
perspective, the feature enables us to distinguish the information transformation
and the state change produced by the execution of the service. The information,
which is presented by the input/output parameters are thought to be reusable,
thus the data values can be duplicated for the execution of multiple services.
Contrarily, the states of the world are changed by the service execution. We
interpret the change as that the old states disappear and the new states are
produced.
To deal with this issue is critical for Web service composition using AI plan-
ning because usually in AI planning, closed world assumption is made, meaning
that if a literal does not exist in the current world, its truth value is considered
to be false. In logic programming this approach is called negation as failure. The
main trouble with the closed world assumption, from Web service construction
perspectives, is that merely with truth literals we cannot express that new in-
formation has been acquired. For instance, one service requester might want to
describe that after sending a message to a Web service, an identity number to
the message will be generated. Thus during later communication the ID could
be used.

50 J. Rao and X. Su
4.3 Rule-Based Planning
Medjahed [18] present a technique to generate composite services from high-
level declarative description. The method uses composability rules to determine
whether two services are composable. The composition approach consists of four
phases. First, the specification phase enables high-level description of the de-
sired compositions using a language called Composite Service Specification Lan-
guage(CSSL). Second, the matchmaking phase uses composability rules to gen-
erate composition plans that conform to service requester’s specifications. The
third phase is selection phase. If more than one plan is generated, in the selec-
tion phase, the service requester selects a plan based on quality of composition
(QoC) parameters (e.g. rank, cost, etc.). The final phase is the generation phase.
A detailed description of the composite service is automatically generated and
presented to the service requester.
Here, we should pay more emphasis on the composability rules because it
is the major issue to define how the plan is generated. The composability rules
consider the syntactic and semantic properties of Web services. Syntactic rules
include the rules for operation modes and the rules for binding protocols of
interacting services. Semantic rules include the following subset: (1) message
composability defines that two Web services are composable only if the output
message of one service is compatible with the input message of another service;
(2) operation semantic composability defines the compatibility between the do-
mains, categories and purposes of two services; (3) qualitative composability
defines the requester’s preferences regarding the quality of operations for the
composite service; and (4) composition soundness considers whether a compo-
sition of services is reasonable. To this end, the authors introduce the notion of
composition templates that define the dependency between the different kinds
of services.
The main contribution of this method is the composability rules, because they
define the possible Web service’s attributes that could be used in service com-
position. Those rules can be used as a guideline for other Web service methods
based on planning.
SWORD [20] is another developer toolkit for building composite Web ser-
vices using rule-based plan generation. SWORD does not deploy the emerg-
ing service-description standards such as WSDL and DAML-S, instead, it uses
Entity-Relation (ER) model to specify the Web services. In SWORD, a service
is modeled by its preconditions and postconditions. They are specified in a world
model that consists of entities and relationships among entities. A Web service
is represented in the form of a Horn rule that denotes the postconditions are
achieve if the preconditions are true. To create a composite service, the service
requester only needs specify the initial and final states for the composite ser-
vice, then the plan generation can be achieved using a rule-based expert system.
Besides the general composition methods, an interesting work done by SWORD
is that the authors give a discussion on that the rule-based chaining can some-
times generate “uncertain” results if a precondition can not uniquely determines
a postcondition. The authors argue that the uncertain results can avoid only

A Survey of Automated Web Service Composition Methods 51
when the preconditions are functionally depending on the postconditions inside
a service. In fact, it may happen in most service composition methods described
in this survey but not all authors explicitly declare it.
4.4 Other AI-Planning Methods
Some other AI planning techniques are proposed for the automatic composition
of Web services. In [26] the SHOP2 planner is applied for automatic composition
of Web services, which are provided with DAML-S descriptions. SHOP2 is an
Hierarchical Task Network(HTN) planner. The authors believe that the concept
of task decomposition in HTN planning is very similar to the concept of compos-
ite process decomposition in DAML-S process ontology. The authors also claim
that the HTN planner is more efficient than other planning language, such as
Golog. In their paper, the authors give a very detail description on the process
of translating DAML-S to SHOP2. In particular, most control constructs can be
expressed by SHOP2 in an explicit way.
Sirin et al [24] present a semi-automatic method for web service composition.
Each time when a user has to select a Web service, all possible services, that
match with the selected service, are presented to the user. The choice of the
possible services is based both on functionalities and non-functional attributes.
The functionalities (parameters) are presented by OWL classes and OWL rea-
soner is applied to match the services. A match is defined between two services
that an output parameter of one service is the same OWL class or subclass of
an input parameter of another service. The OWL inference engine can order
the matched services so that the priority of the matches are lowered when the
distance between the two types in the ontology tree increases. If more than one
match is found, the system filters the services based on the non-functional at-
tributes that are specified by the user as constraints. Only those services who
pass the non-functional constraints can be presented to the service requester.
The idea of semi-automatic service composition is quite interesting because it is
very difficult to capture behavior in sufficient detail and compose the services
in a fully automatic way, especially for the commercial-grade services. Although
the proposed method is simple, it indicates the trend that automatic planner
and human being can work together to generate the composite service for the
user’s request.
4.5 Theorem Proving
Waldinger [25] elaborates an idea for service synthesis by theorem proving. The
approach is based on automated deduction and program synthesis and has its
roots in his earlier work [13]. Initially available services and user requirements
are described in a first-order language, related to classical logic, and then con-
structive proofs are generated with Snark theorem prover. Finally, service com-
position descriptions are extracted from particular proofs.
L¨ammermann [12] applies Structural Synthesis of Program (SSP) for auto-
mated service composition. SSP is a deductive approach to synthesis of programs

52 J. Rao and X. Su
from specifications. The specifications of services only include the structural
properties, i.e. the input/output information. SSP uses propositional variables
as identifiers for input/output parameters and uses intuitionistic propositional
logic for solving the composition problem. The composition is based on the
proofs-as-programs property of intuitionistic logic. It equates the program of
service composition to the problem of proof search. The author also takes takes
advantage of disjunctions in classical logic to describe exceptions, which could
be thrown during service invocation.
Rao et. al. [21, 22] introduces a method for automatic composition of seman-
tic Web services using Linear Logic theorem proving. The method uses semantic
Web service language (DAML-S) for external presentation of web services. And,
internally, the services are presented by extralogical axioms and proofs in Linear
Logic. Linear Logic, as a resource conscious logic, enables people to define at-
tributes of Web services formally (including qualitative and quantitative values
of non-functional attributes). In addition, Linear Logic has close relationship
with π-calculus, which is the formal foundation of many Web service composi-
tion languages. The view of a Linear Logic proof as a π-calculus process was
firstly taken up formally by Abramsky [1], and further elaborated by Bellin and
Scott [3]. The authors attach the π-calculus to the Linear Logic inference rules
in the style of type theory, thus the process model for a composite service pre-
sented by π-calculus can be generated directly from the proof. The authors also
present the subtyping rules that are used for semantic reasoning with LL in-
ference figures. Thus the Linear Logic theorem prover can deal with both the
service specification and the semantic Web information. Unlike other methods
that use non-functional attributes only to filter the generated plan, the authors
consider the non-functional attributes directly in the theorem proving process.
Both service functionalities and non-functional attributes are translated into
propositions in the logical axioms, but the distinguish between the function-
alities and non-functional attributes is enabled by the Linear Logic inference
rules.
5 Conclusion
This paper has aimed to give an overview of recent progress in automatic Web
services composition. At first, we propose a five-step model for Web services
composition process. The composition model consists of service presentation,
translation, process generation, evaluation and execution. Each step requires
different languages, platforms and methods.
In these five steps, we concentrate on the methods of composite Web ser-
vices process generation. We give the introduction and comparition of selected
methods to support this step. The methods are enabled either by workflow re-
search or AI planning. The workflow methods are mostly used in the situation
where the request has already defined the process model, but automatic pro-
gram is required to find the atomic services to fulfill the requirement. The AI
planning methods is used when the requester has no process model but has a

A Survey of Automated Web Service Composition Methods 53
set of constraints and preferences. Hence the process model can be generated
automatically by the program.
Although the different methods provide different level of automation in ser-
vice composition, we can not say the higher automation the better. Because the
Web service environment is highly complex and it is not feasible to generate
everything in an automatic way. Usually, the highly automated methods is suit-
able for generating the implementation skeletons that can be refined into formal
specification. A discussion on this topic is presented by Hull et. al. [11].
Further work will include a more thorough analysis of the field in addition
to practical testing of and experiments with the methods.
References
1. S. Abramsky. Proofs as processes. Theoretical Computer Science, 135(1):5–9, 1994.
2. T. Andrews et al. Business Process Execution Language for Web Services
(BPEL4WS) 1.1. Online: http://www-106.ibm.com/developerworks/webservices/
library/ws-bpel, May 2003.
3. G. Bellin and P. J. Scott. On the pi-calculus and Linear Logic. Theoretical Com-
puter Science, 135(1):11–65, 1994.
4. T. Bellwood et al. Universal Description, Discovery and Integration specification
(UDDI) 3.0. Online: http://uddi.org/pubs/uddi-v3.00-published-20020719.htm.
5. B. Benatallah, M. Dumas, M.-C. Fauvet, and F. Rabhi. Patterns and skeletons
for parallel and distributed computing, chapter Towards Patterns of Web Services
Composition, pages 265–296. Springer-Verlag, 2003.
6. D. Box et al. Simple Object Access Protocol (SOAP) 1.1. Online:
http://www.w3.org/TR/SOAP/, 2001.
7. F. Casati, S. Ilnicki, and L. Jin. Adaptive and dynamic service composition in
EFlow. In P roceedings of 12th International Conference on Advanced Information
Systems Engineering(CAiSE), Stockholm, Sweden, June 2000. Springer Verlag.
8. F. Casati, M. Sayal, and M.-C. Shan. Developing e-services for composing e-
services. In Proceedings of 13th International Conference on Advanced Information
Systems Engineering(CAiSE), Interlaken, Switzerland, June 2001. Springer Verlag.
9. R. Chinnici et al. Web Services Description Language (WSDL) 1.2. Online:
http://www.w3.org/TR/wsdl/.
10. B. N. Grosof, I. Horrocks, R. Volz, and S. Decker. Description Logic programs:
Combining logic programs with Description Logic. In Proceedings of the 12th In-
ternational Conference on the World Wide Web (WWW 2003, Budapest, Hungary,
2003.
11. R. Hull, M. Benedikt, V. Christophides, and J. Su. E-service: A look behind the
curtain. In P roceedings of the 22nd ACM SIGACT-SIGMOD-SIGART Symposium
on Principles of Database Systems (PODS), San Diego, USA, June 2003.
12. S. L¨ammermann. Runtime Service Composition via Logic-Based Program Syn-
thesis. PhD thesis, Department of Microelectronics and Information Technology,
Royal Institute of Technology, June 2002.
13. Z. Manna and R. J. Waldinger. A deductive approach to program synthesis. ACM
Transactions on Programming Languages and Systems, 2(1):90–121, 1980.
14. D. Martin et al. DAML-S(and OWL-S) 0.9 draft release. Online: http://
www.daml.org/services/daml-s/0.9/, May 2003.

54 J. Rao and X. Su
15. D. McDermott. Estimated-regression planning for interactions with Web services.
In Proceedings of the 6th International Conference on AI Planning and Scheduling,
Toulouse, France, 2002. AAAI Press.
16. S. McIlraith and T. C. Son. Adapting Golog for composition of Semantic Web
services. In Proceedings of the 8th International Conference on Know ledge Repre-
sentation and Reasoning(KR2002), Toulouse, France, April 2002.
17. S. McIlraith, T. C. Son, and H. Zeng. Semantic Web services. IEEE Intelligent
Systems, 16(2):46–53, March/April 2001.
18. B. Medjahed, A. Bouguettaya, and A. K. Elmagarmid. Composing Web services
on the Semantic Web. The VLDB Journal, 12(4), November 2003.
19. S. Narayanan and S. McIlraith. Simulation, verification and automated compo-
sition of Web service. In Proceedings of the 11th International World Wide Web
Conference, Honolulu, Hawaii, USA, May 2002. ACM. presentation available at
http://www2002.org/presentations/narayanan.pdf.
20. S. R. Ponnekanti and A. Fox. SWORD: A developer toolkit for Web service com-
position. In Proceedings of the 11th World Wide Web Conference, Honolulu, HI,
USA, 2002.
21. J. Rao, P. K¨ungas, and M. Matskin. Application of Linear Logic to Web service
composition. In Proceedings of the 1st International Conference on Web Services,
Las Vegas, USA, June 2003.
22. J. Rao, P. K¨ungas, and M. Matskin. Logic-based Web services composition: from
service description to process model. In Proceedings of the 2004 International
Conference on Web Services, San Diego, USA, July 2004. IEEE.
23. H. Schuster, D. Georgakopoulos, A. Cichocki, and D. Baker. Modeling and com-
posing service-based and reference process-based multi-enterprise processes. In
Proceeding of 12th International Conference on Advanced Information Systems En-
gineering (CAiSE), Stockholm, Sweden, June 2000. Springer Verlag.
24. E. Sirin, J. Hendler, and B. Parsia. Semi-automatic composition of Web services
using semantic descriptions. In Proceedings of Web Services: Modeling, Architecture
and Infrastructure workshop in conjunction with ICEIS2003, 2002.
25. R. Waldinger. Web agents cooperating deductively. In Proceedings of FAABS 2000,
Greenbelt, MD, USA, April 5–7, 2000, volume 1871 of Lecture Notes in Computer
Science, pages 250–262. Springer-Verlag, 2001.
26. D. Wu, E. Sirin, J. Hendler, D. Nau, and B. Parsia. Automatic Web services
composition using SHOP2. In Workshop on Planning for Web Services, Trento,
Italy, June 2003.