Comparative Assessment of Semantic Relatedness methods on DBpedia
Description
These data refer to the benchmarking of 10 methods for computing semantic similarity applied to 14 golden datasets. The 10 addressed methods are: - Wikipedia Link-based Measure (WLM ) [Witten and Milne, 2008]; - Linked Open Data Description Overlap-based approach (LODDO) [Zhou et al. 2012] - Linked Data Semantic Distance (LDSD) [Passant, 2010] - Linked Data Semantic Distance with Global Normalization (PLDSDGN) [Piao and Breslin, 2016] - Propagated Linked Data Semantic Distance (PLDSD) [Alfarhood et al. 2017] - Information Content-based approach [Schuhmacher and Ponzetto, 2014] - REWOrD [Pirrò, 2012] - Exclusivity-based [Hulpuş et al 2015] - ASRMP [El vaigh et al. 2020] - Proximity-based [Leal, 2013] The 14 golden datasets are: Atlasify240, B0, B1, GM30, MTurk287, WRG252, R122, RG65, MC30, KORE(composed by KORE-IT, KORE-HW, KORE-VG, KORE-TV, KORE-CN) In the experimentation, two DBpedia knowledge graphs have been considered, i.e., with and without the dbo:wikiPageWikiLink links. In addition, we also experimented the 10 methods on pairs of common nouns and pairs of proper nouns, separately, extracted from the 14 datasets.
Files
Institutions
- Consiglio Nazionale delle Ricerche