Interlinking RDF Data in Different Languages

The Semantic Web provides technologies such as the Resource Description Framework (RDF) for representing data on the web. The number of data sets published in RDF is growing rapidly. To connect a variety of data across RDF data sets, they should be interlinked. However, resources can be described in different natural languages. Such publishers as the French National Library [1], the Spanish National Library [2], the National British Museum make their data available using RDF model in their own language. There also exist encyclopedias in RDF: DBpedia in multilingual versions is a structured information from Wikipedia; XLore database [3] is an effort to publish the Chinese encyclopedias (Baidu Baike and Hudong Baike) in RDF. The Europeana Project aims at bringing together descriptions of cultural artifacts from European cultural institutions. This is done by harvesting the metadata of its data providers. The descriptions of these artifacts can be in different languages. The importance of tackling multilingualism in the semantic web has been highlighted in [4]. Problem description. Given two RDF data sets with resources described in different languages, the same entity represented in different data sets has to be identified. At the instance level, the values of properties are in different languages, which makes it harder to merge data about the same entity from different sources. The goal of our research is to identify the same entities across multilingual RDF data sets and link them by owl:sameAs links. For this purpose, we are developing an approach which represents RDF entities as text documents and then compare them. We apply standard Natural Language Processing (NLP) techniques (document preprocessing, term weights, similarity measures) on our data. We particularly explore two strategies presented below.