YAGO4-LP

YAGO-LP datasets

Link prediction datasets based on YAGO4 with the following files in YAGO4-22k/, YAGO4-44k/, YAGO4-77k/, YAGO4-27k/, and YAGO4-19k/:

• rel2id.txt: tab-separated file in which each line associates a relation URI to its equivalent ID
• rel2id.pkl: pickle file containing a dictionary associating a relation URI to its equivalent ID
• id2rel.txt: tab-separated file in which each line associates a relation ID to its equivalent URI
• id2rel.pkl: pickle file containing a dictionary associating a relation ID to its equivalent URI
• ent2id.txt: tab-separated file in which each line associates an entity URI to its equivalent ID
• ent2id.pkl: pickle file containing a dictionary associating an entity URI to its equivalent ID
• id2ent.txt: tab-separated file in which each line associates an entity ID to its equivalent URI
• id2ent.pkl: pickle file containing a dictionary associating an entity ID to its equivalent URI
• class2id.txt: tab-separated file in which each line associates a class URI to its equivalent ID
• class2id.pkl: pickle file containing a dictionary associating a class URI to its equivalent ID
• id2class.txt: tab-separated file in which each line associates a class ID to its equivalent URI
• id2class.pkl: pickle file containing a dictionary associating a class ID to its equivalent URI
• train.txt: train set in which each line is of the form <subjectURI> <relationURI> <objectURI> .
• train.pkl: pickle file containing a list in which each element is a tuple (subjectURI, relationURI, objectURI)
• valid.txt: valid set in which each line is of the form <subjectURI> <relationURI> <objectURI> .
• valid.pkl: pickle file containing a list in which each element is a tuple (subjectURI, relationURI, objectURI)
• test.txt: test set in which each line is of the form <subjectURI> <relationURI> <objectURI> .
• test.pkl: pickle file containing a list in which each element is a tuple (subjectURI, relationURI, objectURI)
• trainIDs.txt: train set in which each line is tab-separated and of the form subjectID relationID objectID
• trainIDs.pkl: pickle file containing a list in which each element is a tuple (subjectID, relationID, objectID)
• validIDs.txt: valid set in which each line is tab-separated and of the form subjectID relationID objectID
• validIDs.pkl: pickle file containing a list in which each element is a tuple (subjectID, relationID, objectID)
• testIDs.txt: test set in which each line is tab-separated and of the form subjectID relationID objectID
• testIDs.pkl: pickle file containing a list in which each element is a tuple (subjectID, relationID, objectID)
• ent2classes.txt: tab-separated file in which each line associates an entity URI to the URI of one of its instantiated classes. Note that instantiated classes are computed taking into account the transitive closure of the instantiation and subsumption relations.
• ent2classes.pkl: pickle file containing a dictionary associating an entity URI to the list of the URIs of its instantiated classes. Note that instantiated classes are computed taking into account the transitive closure of the instantiation and subsumption relations.
• entID2classIDs.txt: tab-separated file in which each line associates an entity ID to the ID of one of its instantiated classes. Note that instantiated classes are computed taking into account the transitive closure of the instantiation and subsumption relations.
• entID2classIDs.pkl: pickle file containing a dictionary associating an entity ID to the list of the IDs of its instantiated classes. Note that instantiated classes are computed taking into account the transitive closure of the instantiation and subsumption relations.
• class2entities.txt: tab-separated file in which each line associates a class URI to the URI of one of its instances. Note that instances are computed taking into account the transitive closure of the instantiation and subsumption relations.
• class2entities.pkl: pickle file containing a dictionary associating a class URI to the list of the URIs of its instances. Note that instances are computed taking into account the transitive closure of the instantiation and subsumption relations.
• classID2entIDs.txt: tab-separated file in which each line associates a class ID to the ID of one of its instances. Note that instances are computed taking into account the transitive closure of the instantiation and subsumption relations.
• classID2entIDs.pkl: pickle file containing a dictionary associating a class ID to the list of the IDs of its instances. Note that instances are computed taking into account the transitive closure of the instantiation and subsumption relations.
• r2dom.txt: tab-separated file in which each line associates a relation URI to its domain URI
• r2dom.pkl: pickle file containing a dictionary associating a relation UI to its domain URI
• rID2domID.txt: tab-separated file in which each line associates a relation ID to its domain ID
• rID2domID.pkl: pickle file containing a dictionary associating a relation ID to its domain ID
• r2range.txt: tab-separated file in which each line associates a relation URI to its range URI
• r2range.pkl: pickle file containing a dictionary associating a relation URI to its range URI
• rID2rangeID.txt: tab-separated file in which each line associates a relation ID to its range ID
• rID2rangeID.pkl: pickle file containing a dictionary associating a relation ID to its range ID
• class2allsuperclasses.txt: tab-separated file in which each line associates a class URI to the URI of one of all its superclasses (taking into account the transitive closure of the subsumption relation)
• class2allsuperclasses.pkl: pickle file containing a dictionary associating a class URI to the list of the URIs of all its superclasses (taking into account the transitive closure of the subsumption relation)
• classID2allsuperclassIDs.txt: tab-separated file in which each line associates a class ID to the ID of one of all its superclasses (taking into account the transitive closure of the subsumption relation)
• classID2allsuperclassIDs.pkl: pickle file containing a dictionary associating a class ID to the list of the IDs of all its superclasses (taking into account the transitive closure of the subsumption relation)
• class2directsuperclasses.txt: tab-separated file in which each line associates a class URI to the URI of one of its direct superclasses
• class2directsuperclasses.pkl: pickle file containing a dictionary associating a class URI to the list of the URIs of its direct superclasses
• classID2directsuperclassIDs.txt: tab-separated file in which each line associates a class ID to the ID of one of its direct superclasses
• classID2directsuperclassIDs.pkl: pickle file containing a dictionary associating a class ID to the list of the IDs of its direct superclasses
• statistics.md: a file containing statistics for the dataset
• additonal_rels.csv: a CSV file indicating which relations were considered when querying and cleaning additional triples to constitute the train set
• yago-relations.csv: a CSV file containing YAGO relations with their domain, range, number of triples and whether they are considered in the valid/test set