Music to Cultural Heritage - Extraction Software A software acting as a prototype for a semantic recommender able to connect songs with cultural heritage objects and agents starting from a single song's identifier.
This study introduces a dual-component solution consisting of the MuCH-Ontology* and the MuCH-ExSoftware. These two components aim to understand the feasibility of a semantic recommender that starts from a song to produce multi-domain suggestions with comprehensible explanations. Even though diverse techniques including the use of large Language Models and other models in the realm of artificial intelligence were leveraged, the resulting prototype still maintains the process transparent and traceable. The prototype also involves external knowledge bases for entity disambiguation and reconciliation, ensuring further transparency and producing queryable knowledge graphs via SPARQL.
The ontology on which the software relies for creating and populating the knowledge graphs can be found here.