Ground Truth for Entity Relatedness Problem over DBpedia datasets

The entity relatedness problem refers to the question of exploring a knowledge base, represented as an RDF graph, to discover and understand how two entities are connected. More precisely, this problem can be defined as: “Given an RDF graph 'G' and a pair of entities 'a' and 'b', represented in 'G', compute the paths in 'G' from 'a' to 'b' that best describe the connectivity between them”.This dataset supports the evaluation of approaches that address the entity relatedness problem and contains a total of 240 ranked lists with 50 relationship paths each between entity pairs in two familiar domains, music and movies, in two subsets of the DBpedia that we called DBpedia21M and DBpedia45M. Specifically, we extracted data from the following two publicly available subsets of the English DBpedia corpus to form our two knowledge bases:1. mappingbased-objects: https://downloads.dbpedia.org/repo/dbpedia/mappings/mappingbased-objects/2021.03.01/mappingbased-objects\_lang=en.ttl.bz22. infobox-properties: https://downloads.dbpedia.org/repo/dbpedia/generic/infobox-properties/2021.03.01/infobox-properties\_lang=en.ttl.bz2 DBpedia21M contains the statements in the mappingbased-objects dataset, and DBpedia45M contains the union of the statements in mappingbased-objects and in infobox-properties. In both cases, we exclude statements involving literals or blank nodes.For each dataset (DBpedia21M and DBpedia45M), the ground truth contains 120 ranked lists with 50 relationship paths each. Each list corresponds to the most relevant paths between one of the 20 entity pairs, 10 pairs from the music domain and 10 from the movie domain, found using different path search strategies.A path search strategy consists of an entity similarity measure and a path ranking measure. The ground truth was created using the following 6 strategies:1. Jaccard Index & Predicate Frequency Inverse Triple Frequency (PF-ITF)2. Jaccard Index & Exclusivity-based Relatedness (EBR)3. Jaccard Index & Pointwise Mutual Information (PMI)4. Wikipedia Link-based Measure (WLM) & PF-ITF5. WLM & EBR6. WLM & PMIThe filename of a file that contains the ranked list of 50 relationship paths between a pair of entities has the following format:[Dataset].[EntityPairID].[SearchStrategyID].[Entity1-Entity2].txtExample 1: DBpedia21M.1.2.Michael_Jackson-Whitney_Houston.txtExample 2: DBpedia45M.27.4.Paul_Newman-Joanne_Woodward.txtThe file in Example 1 contains the top-50 most relevant paths between Michael Jackson and Whitney Houston in DBpedia21M using the search strategy number 2 (Jaccard Index & EBR)The file in Example 2 contains the top-50 most relevant paths between Paul Newman and Joanne Woodward in DBpedia45M using the search strategy number 4 (WLM & PF-ITF)The data is splitted into 2 files, one for each dataset and compressed in .zip format:DBpedia21M.GT.zip: contains 180 .txt files representing the ranked lists of relationship paths between entity pairs in DBpedia21M dataset. DBpedia45M.GT.zip: contains 180 .txt files representing the ranked lists of relationship paths between entity pairs in DBpedia45M dataset.