Flabase: towards the creation of a flamenco music knowledge base

Abstract

Online information about flamenco music is scattered overdifferent sites and knowledge bases. Unfortunately, thereis no common repository that indexes all these data. Inthis work, information related to flamenco music is gath-ered from general knowledge bases (e.g., Wikipedia, DB-pedia), music encyclopedias (e.g., MusicBrainz), and spe-cialized flamenco websites, and is then integrated into anew knowledge base called FlaBase. As resources fromdifferent data sources do not share common identifiers, aprocess of pair-wise entity resolution has been performed.FlaBase contains information about 1,174 artists, 76pa-los(flamenco genres), 2,913 albums, 14,078 tracks, and771 Andalusian locations. It is freely available in RDF andJSON formats. In addition, a method for entity recognitionand disambiguation for FlaBase has been created. The sys-tem can recognize and disambiguate FlaBase entity refer-ences in Spanish texts with an f-measure value of 0.77. Weapplied it to biographical texts present in Flabase. By usingthe extracted information, the knowledge base is populatedwith relevant information and a semantic graph is createdconnecting the entities of FlaBase. Artists relevance is thencomputed over the graph and evaluated according to a fla-menco expert criteria. Accuracy of results shows a highdegree of quality and completeness of the knowledge base.