The increasing importance and use of Web services have resulted in a number of e orts targeted at automating Web service discovery and composition based on semantic descriptions of their properties. However, the progress in the automation of Web service discovery is still held back by the fact that the description of Web services in terms of semantic metadata is still mainly manually. This Ph.D. thesis addresses this problem by developing an approach for the acquisition and management of semantic Web service descriptions in order to facilitate e cient service discovery and composition. Speci cally, this involves the collection of information about a Web service, the acquisition of semantic descriptions based on the collected information, and the structured storage of the generated semantic descriptions.
A clearly developing trend in the recent years is towards exposing the functionalities provided by existing software components in the form of services and facilitating in this way software reuse and added value through service composition. This development is strongly supported by the Web, which enables ubiquitous access via a set of Web standards and protocols to software components residing on di erent platforms. As a result, Web services are seen increasingly as the basic construct for the development of rapid, low-cost and easy-to-compose distributed applications in heterogeneous environments.
When it comes to nding appropriate services and composing distributed applications, current technologies still require a large amount of human interaction. This results in the restricted number of use cases for service integration and the limited scalability of solutions involving manual activities in the process of service discovery and compositions. In order to address these problems, the idea of supplementing Web services with a semantic description of their functionality, that could facilitate their discovery and integration, has been developed. However, the information, on which these semantic descriptions are based, is rarely discussed. In addition, there is still no established method for acquiring semantic Web service descriptions, which not only describe the functionality but also specify functional and non-functional properties in a way that supports automatic matchmaking processes. Therefore, there is the strong need of developing an automatic approach, which enables the acquisition and management of semantic Web Service descriptions with the goal of supporting e cient service discovery and composition.
1
Technical Service
Descriptions Natural Language
Descriptions User-provided Tags, Ratings, Comments 2
Extraction of Functional and Non-Functional Properties 3 Semantic Web Service Descriptions This Ph.D. thesis targets to explore the problem of semi-automatically acquiring and publishing semantic descriptions for Web services and aims to use Semantic Web techniques to develop a methodology to enhance the situation. Figure 1 shows the three main challenges connected with the acquisition and management of semantic Web Service descriptions. 1. How can information about Web services be collected? To answer this question, the relevant sources of information have to be identi ed, including not only technical and text descriptions, but also information inspired by Web 2.0 such as user tags, ratings and comments. After this, automatic mechanisms for data collection from the Web have to be developed. 2. How can semantic Web service descriptions be automatically acquired? This question is in the focus of the Ph.D. thesis and is addressed by exploring existing methods for metadata acquisition and identifying their advantages and disadvantages. The main goal is to develop a method, which semi-automatically acquires metadata by extracting functional and nonfunctional service properties in order to facilitate discovery and composition. 3. How can semantic Web service descriptions be stored? The acquired semantic descriptions have to be stored in a way that facilitates the manual as well as the automatic matchmaking processes. It is important to employ the metadata's syntactical or semantical interconnections in order to enable indexing and e cient service search. 2
There is already some work done related to the automatic creation of semantic
Web service descriptions. In particular, there are two main areas of research: the
acquisition of a suitable Web service domain ontology and the actual process of
annotating Web services. Sabou et al. [
Focusing on the Web service annotation task, Patil et al. [
There are a number of state of the art Web service repositories including
UDDI, Bindingpoint, .NET XML Web Services Repertory, WebserviceX.NET,
Web Service List, Xmethods and SalCental. An overview of these repositories
presented in [
A number of Web service tasks can be automated by using semantic descriptions. In particular, service o ers and requests can be matched automatically. However, semantic descriptions still have to be manually generated. The major goal of this Ph.D. thesis is to develop an semi-automatic method for the acquisition of semantic descriptions of Web services, based on metadata and tags extracted from the Web. The resulting semantic descriptions will be stored in an online semantic Web service descriptions repository, which can be used for both manual and automatic service discovery. The acquisition of semantic descriptions facilitates matchmaking approaches and can reduce the required human interaction to a minimum. Especially in use cases, where Web services are used as building blocks in distributed applications or where Web services provide functionality integrated in business processes, the automated deriving of semantic description plays a key role.
There are three main contributions, which are to be archived. First, the automatic collection of information describing a Web service is facilitated. Second, semantic descriptions of Web services are semi-automatically acquired from the collected information, in the form of semantic annotations. Finally, Web service semantic annotations are stored in an online repository, which provides semantic indexing to allow for e cient queries and support matchmaking processes. Each of these contributions is described in detail in the following sections.
Web Service Crawlers are developed to support the automatic collection of information about Web services. Web service descriptions are usually given in a standardized form such as WSDL. Unfortunately, WSDL provides mainly technical description of services and this information is insu cient for automatic service discovery because it presumes increased human interaction. Therefore, it is necessary to automatically collect additional information. A Web service crawler can e ectively be used to perform this task. In particular, this involves the identi cation of relevant sources and types of Web service information. Sources include natural language text descriptions, for example, documentation on provider Web-sites, and technical Web service descriptions such as WSDL, or WSMO variants. In addition, sources of information inspired by Web 2.0 such as user tags, ratings, and community inputs are also considered. The de nition of the Web service crawlers is based on semantic attributes for service description, based on information needed for service discovery and composition. This involves mainly the speci cation of functional and non-functional Web service properties.
Web Service Semantic Annotations comprise the acquired semantic Web
services descriptions, which are based on the data collected by the Web service
crawlers. The main focus of this Ph.D. thesis is on developing a mechanism
for the acquisition of semantic descriptions from the collected data, based on a
determined formalism for the speci cation of semantic information. This
mechanism is developed by using existing techniques for data mining and ontology
learning and thereby, automatically constructing service annotations. In
addition, the formalism, chosen as the basis for the semantic annotation, annotates
functional and non-functional properties of Web services in a unifying way. The
approach described in [
Semantic Indexing is necessary in order to facilitate e cient queries on top of the semantic annotations. The structured storage of the acquired semantic annotations is targeted to improve performance of automatic matchmaking processes and speed up manual search. During semantic indexing, the collection of Web service semantic annotations will be analyzed and organized in a way that allows the e cient answering of expressive queries. 4
The planned evaluation consists of three main steps. First, the developed mechanism for data collection is evaluated by determining a xed domain of source data and assessing how much data was correctly retrieved. This evaluation can easily be performed by using the well-established recall/precision metrics. Second, semantic annotations evaluation assesses the performance of extracting Web service properties relevant for matchmaking from the corpus of collected data. Naturally, the quality of semantic annotation extraction has a direct in uence on the quality of the service discovery and composition, which are based on these semantic annotations. This evaluation can be done by comparing the Web service functional and non-functional properties present in the collected data to the ones present in the acquired semantic descriptions. Third, the chosen structure for storage of Web service semantic annotations is evaluated by comparing the performance of matchmaking algorithms ran on the repository to the performance of these matchmaking algorithms ran on the same semantic annotations but in unstructured list collection. This evaluation will point out what improvements the developed repository structure brings in means of search performance. Finally, the best evaluation for the developed approach would be the making public of the semantic annotations repository so that other researchers can use it to test their discovery and composition algorithms. 5
The work plan is divided into three stages, each stage marking one year of the Ph.D. studies. Initialization State of the art report on formalisms for Web service semantic descriptions, in the context of service discovery (M6). State of the art report on data-mining techniques and ontology learning (M6). State of the art report on possible structures for semantic annotations storage (M6). Identied sources of Web service information (M12). Identi ed formalism for semantic annotations (M12). Development De nition of semantic attributes for the Web service crawlers. First Web service crawlers prototype (M18). Identi ed structure for annotations storage (M18). Identi ed data-mining and ontology learning approach (M24). First prototype implementation of acquisition of semantic annotations (M24). Evaluation of the rst prototypes (M24). Re nement First prototype of semantic annotations repository (M30). Re ned methodology based on the rst prototypes (M30). Re ned implementation (M36). Evaluation (M36).