A hybrid approach for artwork recommendation

IGNACIO GATTI; JORGE ANDRES DIAZ PACE; SILVIA SCHIAFFINO

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE

PERGAMON-ELSEVIER SCIENCE LTD

Lugar: Amsterdam; Año: 2023 vol. 126

0952-1976

In this article, we present an approach for artwork recommendation. Museums usually exhibit thousands of artworks. Moreover, the museums has their collection online; therefore visitors can contemplate the artworks from any digital device. However, in both cases, artworks are not randomly distributed. The curators are responsible for the organization of an artwork collection seeking a delicate balance between emotion and reason. Despite that, the visitor´s interests are likely to match only a subset of the artworks. This problem can be seen as a recommendation problem. We hypothesize that the artwork nature is a fundamental aspect when designing a Recommender System in the art domain. Our proposal is a hybrid recommendation approach thatcombines deep autoencoders with an influence social graph in order to complement visual aspects and context of the artworks in recommendations. In order to verify our hypothesis, we performed a case-study with a group of experts. The results obtained, although preliminary, showed that the proposed approach is an interesting alternativeto recommend artworks.