Cross-lingual Property Alignment for DBpedia ontology using Triple Conceptualization

Property mappings between different lingual datasets are important for linking datasets and constructing an unified multilingual ontology regardless of the language. We conduct a preliminary evaluation in DBpedia datasets using triple-conceptualization approach by suggesting type information of instances in triples represented as vector. We also show how this vector representation of the types helps the instance matching based approach. Our results indicate that type information is beneficial to find semantically corresponding properties in two different languages without translation, but its use remains limited in URI-type triples. This paper is meant as a basis for applying the property mapping between two language datasets to a DBpedia ontology enrichment.