Completing Predicates Based on Alignment Rules from Knowledge Graphs

Knowledge graphs (KGs) are dynamic structures, often shaped by diverse user communities, leading to the emergence of alternative representations for the same concepts. These alternative definitions, while enriching KGs with complementary information, also pose a challenge for downstream tasks by potentially impeding the completeness of the retrieved information. This paper tackles the problem of identifying alternative definitions of predicates within KGs. We present SYRUP, a method designed to uncover conjunctions of predicates that encapsulate the same semantic relationship as a given predicate but offer complementary instances. Through SYRUP, we aim to augment KG completeness by harnessing these alternative representations. To assess the effectiveness of SYRUP, we conduct an empirical study using a benchmark of 60 SPARQL queries over DBpedia, comprising six distinct domains. Our experimental results demonstrate improvements in both the completeness and correctness of query answers, with accuracy levels ranging from 0.73 to 0.95. Furthermore, we make SYRUP openly accessible on GitHub ( https://github.com/SDM-TIB/SYRUP/ ), enabling researchers to replicate our experiments and integrate SYRUP into workflows for KG enhancement.