In this paper, we propose an approach for mapping properties in two RDF datasets between diferent languages, using a triplebased similarity propagation that can be adapted to find potential property matches. This approach does not need any language dependent information during the process, and thus can be applied to arbitrary languages without requiring translation.
Introduction
The proposed approach has two steps: 1) findings the equivalent subject and
object values across datasets at the entity-level, which is represented in the form
triples, that are connected by owl:sameAs links, and then considering the
associated properties to have the potential to be equivalent. 2) Then, using a small
number of identified matches as seeds to exploit the conceptual-level alignments
to identify and estimate semantic relatedness of properties. Often, the
conceptualizations of triples (from instance triples) are eficient in terms of coverage of
alignment, but their result may be dependent on recognizing entities and their
type. The types of an entity may not always be present in the dataset. The
‘similarity flooding approach’ [
Eun-kyung Kim et al.
Experiments: The goal of this experiment is to align language-local properties (i.e., DBpedia Korean property in this case) with the ontological properties of DBpedia in English. Three human annotators aligned 1,000 DBKP to DBOP, if the meaning of two properties was similar. We used the majority vote to determine the correct mapping results. Analysis: The preliminary experiment between the English and the Korean DBpedia has shown that the propagated connectives improve the recall and F1-score measures required to nfid mapping pairs of properties by taking into account instance types in order to discover new mapping candidates. We see this as the initial step towards enhancing multilingualism in Linked Open Data. Acknowledgement This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIP) (No. R0101-15-0054, WiseKB: Big data based self-evolving knowledge base and reasoning platform)