DBpedia FindRelated is a service to retrieve resources that are related to a given seed resource. For each resource retrieved, the distance indicating the relatedness is included. There are two models that can be used for computing relatedness:
- symmetric uses symmetric distances, i.e., the distance between the seed resource and the retrieved resource is the same as vice versa
- asymmetric uses asymmetric distances, i.e., the distance between the seed resource and the retrieved resource is not the same as vice versa
The DBpedia Type Completion Service can be called with three parameters: resource is the DBpedia resource for which to predict types, model is the model to use (symmetric or asymmetric), and threshold is a cutoff threshold for the distance. The result is a list of resources and distances.
Example call: http://wifo5-21.informatik.uni-mannheim.de:8080/DBpediaFindRelated/Service?resource=http://dbpedia.org/resource/Mannheim&model=symmetric&threshold=0.1
The service uses two trained models (in particular, Support Vector Machines), built from datasets of randomly selected, connected pairs of DBpedia resources, and using the Normalized Web Search Engine Distance, computed with Yahoo!, as ground truth. The dataset is available here.
The DBpedia Find Related Service has been developed by Heiko Paulheim, University of Mannheim.
DBpedia Spotlight's Rel8 endpoint: https://github.com/dbpedia-spo[..]-spotlight/wiki/Rel8