- From: Raphael Volz <volz@fzi.de>
- Date: Wed, 12 Dec 2001 19:32:38 +0100
- To: <www-webont-wg@w3.org>
- Cc: <ls3@aifb.uni-karlsruhe.de>
- Message-ID: <BC630F6C39BFAB4CA1B35EFA0C7F217F0556FB@stan.fzi.de>
Hi all ! again, apologies for the late delivery. But travel responsibilities resulted in being offline for the last 3 weeks and I didn't get registered before... Here they are... Best regards, Raphael ----- Karlsruhe Use Cases: 1) Classical Ontobroker and SHOE stuff: Annotating web pages with additional metadata is used to explicitly "type" the information using an ontology. This approach can be (and was successfully) applied to provide querying on a semantic level (instead on a keyword level) and improve query results. Queries essentially operate on data instances instead of full-text This was used in the seminal AIFB KA2portal, and the German GETESS project. The annotations can also be used by automated agents to fulfill tasks. 2) In the first scenario ontologies provide a single, common vocabulary that acts as a communication basis between humans and machines, viz. as a semantic data model for integration of heterogeneous data and as a common business terminology that is used. In the FZI research project "Ontologging" goes beyond this single common vocabulary within a department. A challenging problem is the global knowledge access through different departments supporting different vocabularies, e.g. establishing communication between the finance and the human resource department, on the human and on the machine level. The idea behind this approach is that people use their own ontololgy, but mappings between different ontologies (in our case to other departments) are discovered automatically and proposed to the users within the use of the system. The user may than decide if the mapping should be explicitly represented or neglected. The ontology language must then be able to express (or atleast facilitate) such mappings (i.e. DAML+OIL equivalent can be used to express equivalence of classes / properties). 3) Additionally to this, one could also think of applications that make their business rules or rules that are described in natural language explicit by representing it in machine-understandable way. Think for example of a biological community that does research on "trees" - one could publish research results about whether a certain type of tree grows better with a certain soil, with certain amounts of light, etc. Here, a web ontology language would also need the ability to represent "rules". Then, some researcher could use the ontology to discover the results of other research agencies in goal-directed queries and augment his own "knowledge" with the discovered rules. Also machine agents could use these rules in their automated reasoning enhancing the problem solving capabilites. For example a machine could determine which interval and quantity of water is best to be used when taking care of plants (i.e. while the owner is on travels). 4) In the portal for the EU-supported OntoWeb network of excellence on Ontologies and Knowledge Management ontologies are used very much like in the seminal Ka2Portal that is to provide semantic querying. Additionally ontologies provide a way to syndicate the content published by the members. Here, classes and relations are used to describe structured content types (e.g. Publication, News (RSS), Event, ...) these content types are then gathered from the syndicated web sites to form the content of the portal. For this application reification on a model level, that is on a group of statements, is required to denote who is the author of a certain piece of content. So http://foo/bar might publish some events and these events are displayed within ontoweb.org. This content type is then augmented with author/publisher information (we use DC). Sophisticated publish/review mechanisms exist to control the published contents . 5) Web ontology usage within the EU-funded On-To-Knowledge project (provided by York Sure (sure@aifb.uni-karlsruhe.de)): Case study 1: Organizational Memory Swiss Life (Switzerland) is a large insurance company serving customers around the world. Their vision is to build an organizational memory with an intranet based portal that offers a single entry point to the knowledge space of the company. The case study explores different parts of their intranet: 1. Skills management makes skills of employees explicit. Within the case study existing skill databases and documents (like e.g. personal homepages) will be integrated and expanded. Two aspects are covered by the case study: first, explicit skills allow for an advanced expert search within the intranet. Second, one might explore his future career path by matching his current skill profile vs. job profiles. The skill ontology serves as a shared conceptualisation of relevant skills for profiles. 2. The International Accounting Standards (IAS) document is part of the global Swiss Life Intranet. The content of the document is highly specialized and even trained people hardly find relevant passages, even though there is a division into chapters and sections. Providing sophisticated access to the IAS is the second goal of this case study. Case study 2: Call Centre Desktop Support British Telecom (United Kingdom) is a leading company on the telecom market. An important segment of their service portfolio are call centres. High quality information is needed to be able to answer complex questions of customers. Every transaction should emphasize the uniqueness of both the customer and the customer service person, i.e. call centre agents. Therefore agents need to share their knowledge effectively with colleagues so that existing solutions from other agents are available to all agents. In the case study an ontology serves to structure the knowledge space by providing a shared and agreed upon model. The goal of the case study is to enable a shared knowledge space with sophisticated access to existing solutions. Case study 3: Virtual Organisation EnerSearch (Sweden) is a virtual organization founded by a consortium of major energy providers researching new IT-based business strategies and customer services in deregulated markets [10]. Its research affiliates and shareholders are spread over many countries (e.g. US, Sweden, Germany ^Å). Essentially EnerSearch creates knowledge which is then transferred to shareholders and other interested parties. Goal of the case study is to enhance the knowledge transfer to researchers in different disciplines and countries, and to specialists from shareholding companies interested in getting up-to-date information about R&D results on IT in Energy. Ontologies will help to enable a content based search on research topics. An important aspect of the case study is an initial experiment where end-users compare different search methods including traditional keyword-based search, ontology-based querying and ontolgoy-based brwosing. 6) Ontology usage for text / data mining (Hotho et al. - http://www.aifb.uni-karlsruhe.de/WBS/Publ/2001/hothoetal.pdf) Ontologies can be used to bring background into machine learning and explain the results of learning algorithms. For example, text clustering typically involves clustering in a high dimensional space, which appears difficult with regard to virtually all practical settings. In addition, given a particular clustering result it is typically very hard to come up with a good explanation of why the text clusters have been constructed the way they are. We developed views on the high dimensional data basing our selection of features on a heterarchy of classes. The results can be distinguished and explained by the corresponding selection of classes in the ontology. 7) Combination of ontologies and web mining -> Semantic Web mining (provided by Gerd Stumme (stumme@aifb.uni-karlsruhe.de)) The Semantic Web and Web Mining are two fast-developing research areas which have a lot of points of contact. Results of Web Mining can be improved by exploiting the new semantic structures in the web; on the other hand, Web Mining can be used for building the Semantic Web. Ontologies play an important role in this integration task. For Web Mining, the levels from XML and RDF to ontologies and logics are of particular interest. Web Mining applies data mining techniques on the web. Three areas can be distinguished: Web usage mining analyzes user behavior, web structure mining utilizes the hyperlink structure, and web content mining exploits the contents of the documents on the web. Semantic Web Mining aims to brings together research on the different represenation levels of the Semantic Web, and on the different Web Mining approaches. First results on this trend have been presented at the Semantic Web Mining Workshop at the 12th European Conference on Machine Learning (ECML'01)/5th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD'01) in September 3-7, 2001, at Freiburg, Germany <http://semwebmine2001.aifb.uni-karlsruhe.de/> 8) Issues in Semantic Annotation (provided by Siegried Handschuh (handschuh@aifb.uni-karlsruhe.de)) Richly interlinked, machine-understandable data constitutes the basis for the Semantic Web. Annotating web documents is one of the major techniques for creating metadata on the Web. However, annotation tools so far are restricted in their capabilities of providing richly interlinked and truely machine-understandable data. They basically allow the user to annotate with plain text according to a template structure, such as Dublin Core. We here present CREAM (Creating RElational, Annotation-based Metadata), a framework for an annotation environment that allows to construct \emph{relational metadata}, i.e.\ metadata that comprises class instances and relationship instances. These instances are not based on a fix structure, but on a domain ontology. We discuss some of the requirements one has to meet when developing such a framework, e.g.\ the integration of a metadata crawler, inference services, document management and information extraction, and describe its implementation, Ontomat a component-based, ontology-driven tool.
Received on Wednesday, 12 December 2001 13:32:40 UTC