Entity Label Mapping dataset

This is an Entity Label Mapping dataset produced from Wikidata entities. It lists labels and aliases recovered from the knowledge graph in ten pre-selected languages. This data has been used and described in our paper, currently under anonymous review. The files, alongside their descriptions, are: all_entity_labels.json: All extracted labels of entities identified as either Person, Organisation, or Place within Wikidata, on all ten pre-selected languages. all_entity_labels_criteria.json: A filtered version of the previous file, containing only those entities deemed rich in multi-lingual labels (within the aforementioned ten languages). entity_label_mapping_benchmark_dataset.json: The benchmark dataset for the entity label mapping task, as described in the paper. entity_label_mapping_benchmark_dataset_sample_scored.json: A sample of the previous document that has been scored for label similarity according to a number of approaches. entity_label_mapping_ground_truth.json: A smaller sample of the previous document, this time also manually annotated with ground-truth labels, and with similarity scores converted into best-match labels.