Universal Schemas Using Shortest Dependency Paths for Free Word Order Languages

Abstract

Universal schemas are a remarkable approach for solving relation extraction, in which new facts are extracted by jointly learning latent feature vectors of entity pairs and relation types through matrix factorization models. However, in free word order languages where surface form predicates do not fully reveal the characteristics of a sentence, universal schemas cannot be constructed in the same manner. Therefore, in this study, we introduce a novel expansion of universal schemas, dependency-path-based universal schemas. Our model uses shortest dependency paths and entity types instead of surface form predicates. For verification of our model, we constructed and evaluated a universal schema in Korean with a combination of Korean DBpedia and Korean Wikipedia distant supervision data.