Abstract We present our vision of a Semantic Web en-hanced desktop environment, the Semantic Desktop. Common desktop resources like MS-Outlook contacts or Mozilla bookmarks are treated as Semantic Web Resources. The Gnowsis system is our open source project that implements this vision to be used and ex-tended by developers. 1 Introduction The Semantic Web is still in its beginning. We have to push forward and start building useful applications. The Semantic Desktop approach does not focus on or-ganizational or worldwide systems, it’s focus is to or-ganize data on personal computers. The use of Seman-tic Web technology in desktop applications will create a benefit for users. When users benefit from it, the technology will be used and the Semantic Web will grow. Tim Berners-Lee tells us from time to time to start writing applications [Berners-Lee 2003]. On a modern personal computer there are thousands of documents, emails, photos and music songs stored. Their number increases as users download new files, send and receive emails, socialize and exchange con-tact information. A normal application can handle only part of the whole data. All these information items can be treated as Semantic Web resources. Existing technology and approaches from the Semantic Web are then used on these desktop resources. Metadata of as well as relationships between all the resources on a computer are represented in the Resource Description Framework (RDF). Data from different applications are linked using triples. A Se-mantic Browser is used to visualize the information and allow users to benefit. This results in a Semantic Web enhanced Desktop – the Semantic Desktop. Ap-plications still handle only their part of the whole – and access the other parts through RDF. There are different approaches in this topic. Haystack [Quan et al., 2003] is similar to Gnowsis. Haystack is also a framework to build RDF applications, having its own user interface and RDF engine. In the Gnowsis project, we decided to build an architecture to integrate existing applications instead of replacing them. In to-day’s software projects it is a common demand that new software integrates with Microsoft’s Outlook or popular tools like the Mozilla suite. We created a framework to integrate different applications. Using this data integration framework, it is possible to extract information on the fly from common applications. De-velopers can also extend the framework and integrate custom data sources. The Semantic Desktop will play a key role in building a global Semantic Web. First users will annotate their personal data, motivated by the benefits of an inte-grated personal information management. Then this annotated data can be published on the web. Web sites don’t have to be annotated with the “add semantic sugar on top” philosophy, instead all re-sources can be annotated starting with the time of creation. Once a file is annotated, it will be a small step to add the annotations when the file is sent through email, posted on a website or stored in an organizational memory. 1 . 1 Everyt h ing is a Resour c e To build such a system, an experimental approach in the Semantic Web field had to be taken. While people still discuss the meaning and interpretation of URIs on the RDF-IG newsgroup and the W3 [Berners-Lee 2002], we tested different approaches and implemented them. All resources on the desktop are identified by a URL, be it a file, an email or a database row. These URLs can be used to annotate the resources using third party applications. The framework parses the URLs to find the resources. There are many ways to exchange RDF data between applications or how to query them from servers. Avail-able implementations are the Joseki Server1 by HP Labs or Sesame2 which both provide HTTP interfaces. We decided to use RDQL3 to find resources and an 1 http://www.joseki.org/ 2 http://www.openrdf.org/ 3 http://www.hpl.hp.com/semweb/rdql.htm Introducing the Gnowsis Semantic Desktop Leo Sauermann, Sven Schwarz DFKI GmbH Knowledge Management Group Kaiserslautern, Germany { leo.sauermann | sven.schwarz }@dfki.deextension of Concise Bounded Descriptions (CBD) [Stickler 2004] to retrieve the RDF data. 1. 2 Im p l e m en t a t i o n The popular RDF framework Jena4 by HP Labs was chosen as the main technical ground to build Gnowsis on. We extended the Jena framework with the Semantic Desktop functions. For different data sources like the file-system or Microsoft Outlook we created adapters, which are accessible as Jena models. We also inte-grated existing adapters, for relational databases we used D2RQ from Chris Bizer [Bizer 2004]. Developers and Researchers can use these Jena Models in their own applications to extract information from the dif-ferent data sources. Developers can access the functionality through XML/RPC and RMI interfaces. The framework is Java based and adapted to Windows. We are working on Linux integration. 2 System Design The architecture will be explained using a typical use case. The system is designed for typical tasks of to-day’s users, using common applications and the popu-lar MS-Windows operating system. 2 . 1 Typi cal Use Case A typical use case is that the user wants to know more information about a music file. Right-clicking the file allows the user to “browse” in his personal Semantic Web. The request is sent to the local Gnowsis server passing the URL of the file and the “browse” command identifier. It parses the URL and identifies it as a file. There are several adapters registered in the system, one handles local files and can extract file-size and name. Another adapter can extract the ID3-Tag information of the MP3 file, like artist and track number. Each adapter is shipped with an RDFS vocabulary describing the possible values that the adapter can extract. The central server queries both adapters about the MP3 file. Each adapter extracts his part of the information.