Publishing Linked Data
This research theme is all about creation, interlinking and deployment of linked data in RDF/XML and RDFa. We investigate RDFiser - services that turn (semi)structured data into RDF - and interlinking methods ranging from automatic interlinking to user-contributed interlinking.
Currently we are mainly working on any23, a library and web service that extracts structured data in RDF format from a variety of Web documents. We use any23 in Sindice to extract structured metadata from Web pages.
Bridging the Semantic Desktop and Linked Data
Contact: Tudor Groza
We have developed an application that implements a three-step process with the goal of enriching scientific publications with semantic metadata. Starting from an empty model, incrementally and driven by the user, the application performs: (i) automatic extraction of shallow metadata (title, authors) from a given publication; (ii) expansion and cleaning of the extracted metadata based on linked data repositories (currently the Semantic Web Dog Food Server and Faceted DBLP); (iii) integration of the linked metadata within the personal information model by means of the Semantic Desktop. The result is two-folded: conceptually, a clean and linked semantic model of the publication, while its integration in the personal space provides a means for weaving linked publication data using the Semantic Desktop and typical KDE desktop applications.
A demo of the complete features of the application can be found at: http://sclippy.semanticauthoring.org/movie/sclippy.htm
We have developed some specialised interlinking modules for the software engineering domain (Java2RDF, SVN2RDF, etc.) and experiment with Silk, a link discovery framework for the Web of Data.