Abstract. Dissemination, an important phase of scientific research, can be seen as a communication process between scientists. They expose and support their findings, while discussing claims stated in related scientific publications. However, due to the increasing number of publications, finding a starting point for such a discussion represents a real challenge. At same time, browsing can also be difficult since the communication spans accross multiple publications on the open Web. In this paper we propose a semantic claim federation infrastructure, named KonneX SALT, as a solution for both issues mentioned above: (i) finding claims in scientific publications, and (ii) providing support for browsing by starting with a claim and then following the links in an argumentation discourse network (ADN) (in our case, by making use of transclusion). In addition, we join the web of linked open data, by linking the metadata contained in KonneX SALT with some of the known repositories of scientific publications. 1

Abstract. With the advancement of technology and the wide adoption of ontologies as knowledge representation formats, in the last decade, a handful of models were proposed for the externalization of the rhetoric and argumentation captured within scientific publications. Conceptually, most of these models share a similar representation form of the scientific publication, i.e. as a series of interconnected elementary knowledge items. The main differences are given by the terminology used, the types of rhetorical and / or argumentation relations connecting the knowledge items and the foundational theories supporting these relations. This paper analyzes the state of the art and provides a concise comparative overview of the five most prominent discourse representation models, with the goal of sketching an unified model for discourse representation. 1

Abstract. Semantic metadata can be considered one of the foundational blocks of the Semantic Web and Desktop. This report describes a solution for automatic metadata extraction from scientific publications, published as PDF documents. The proposed algorithms follow a low-level document engineering approach, by combining mining and analysis of the publications ’ text based on its formatting style and font information. We evaluate them and compare their performance to other similar approaches. In addition, we present a sample application that represent the use-case for the metadata extraction algorithms.

and other research outputs A short survey of discourse representation models Conference Item