What are Semantic Annotations? Eyal Oren1, Knud Hinnerk M��oller1, Simon Scerri1, Siegfried Handschuh1, and Michael Sintek2 1 Digital Enterprise Research Institute, National University of Ireland, Galway {eyal.oren, knud.moeller, simon.scerri, siegfried.handschuh}@deri.org 2 German Research Center for Artificial Intelligence (DFKI) sintek@dfki.uni-kl.de Abstract. Annotations of Web resources can be created using tradi- tional document annotation tools or more recent approaches such as semantic wikis, semantic blogs and collaborative tagging. Currently no unified model exists for all these different kinds of annotations, making it difficult both to compare and assess annotation tools and to integrate the various kinds of annotation data. We analyse annotations in various domains and present a unified formal model for semantic annotations. We evaluate existing annotation tools from these different domains and show how to map the data these tools produce onto our formal model, thus allowing to access and represent this data in a unified way. 1 Introduction The Semantic Web enables machines to interpret, combine, and use data on the Web. Whereas the current ���eye-ball��� Web is only understandable for humans, the Semantic Web can be used by computers as well. The basis for the Semantic Web are computer-understandable descriptions of resources. We can create such descriptions by annotating resources with metadata, resulting in ���annotations��� about that resource. Metaphorically, we can see URIs3 as the ���atoms��� of the Semantic Web and semantic annotations as the ���molecules���. The Semantic Web is about shared terminology, achieved through consistent use of URIs. Annotations create a re- lationship between URIs and build up a network of data. Several tools and paradigms exist to create annotations of Web resources, both manually, semi-automatically and fully automatically. Traditionally the domain of (shallow) linguistic analysis in document annotation, we observe the current trend of tagging, semantic blogs and semantic wikis as forms of annotations as well. Currently, no unified formal model exists to describe these different kinds This material is based upon works supported by the Science Foundation Ireland under Grants No. SFI/02/CE1/I131 and SFI/04/BR/CS0694 and by the European Commission under the Nepomuk project FP6-027705. 3 http://www.w3.org/Addressing/

of annotations. The lack of such a model limits the comparison of annotation tools and complicates the integration of annotation data from multiple sources. Bringing all annotations into a common RDF4 model would partially solve the data integration. However, to map annotations into RDF we need to first understand their similarities and differences: we need a common model of an- notations, and we need to unify the terminology and modelling choices of the various tools. In this paper, we analyse the notion of ���annotation��� in the context of the Semantic Web: we investigate annotations in different domains (document an- notation, tagging, semantic wikis and semantic blogs), we define a formal and unified model of annotations, derive categorising dimensions and evaluate exist- ing annotation tools according to these dimensions. Finally we show how existing annotation tools can be mapped to our unified formal model, allowing data in- teroperability and integration. Our work is based on several earlier approaches to understand, classify and compare document annotations such as [3,7,17,20]. We align and extend these works with new paradigms such as Semantic Wikis and Semantic Blogs and introduce a formal model for annotations. Our contributions are: (i) a formal model of annotations with characterising dimensions, (ii) a clear characterisation of annotation tools, and (iii) a unified model for semantic annotations. 2 Creating annotations The term ���annotation��� implies, very generally speaking, to attach data to some other piece of data. During the course of this paper, we will elaborate on this rather simple statement with respect to a number of different domains. The following sections will give a short introduction to each of these domains and specify the role of annotations for each domain. 2.1 Document annotations The traditional domain of document annotation covers the annotation of arbi- trary textual documents, or parts of them. Annotations can be manual (per- formed by one or more people), semi-automatic (based on automatic sugges- tions), or fully automatic. Manual annotation tools allow users to add annotations to web pages or other resources, and share these with others. An example annotation would relate the text ���Paris��� to an ontology, identifying it as a city and as capital of France. Auto- matic tools can perform similar annotations (such as named-entity recognition) without manual intervention. 4 http://w3.org/RDF