Ontology-Based Identification of Music for Places

Abstract

Esta es la versión de autor de la comunicación de congreso publicada en: This is an author produced version of a paper published in: Abstract: Place is a notion closely linked with the wealth of human experience, and invested by values, attitudes, and cultural influences. In particular, many places are strongly linked to music, which contributes to shaping the perception and the meaning of a place. In this paper we propose a computational approach for identifying musicians and music suited for a place of interest (POI). We present a knowledge-based framework built upon the DBpedia ontology, and a graph-based algorithm that scores musicians with respect to their semantic relatedness to a POI and suggests the top scoring ones. We found that users appreciate and judge as valuable the musician suggestions generated by the proposed approach. Moreover, users perceived compositions of the suggested musicians as suited for the POIs.

Utilizing transfer learning for in-domain collaborative filtering

“…The idea led to further research based on exploring user/item domain correlations via latent factors learning [5,10,16]. Other examples include: estimating domain correlations based on tag similarities [12], linking concepts and items based on ontology [44], and linking user personality types with domain-dependent preferences based on association rules [45].…”

Cross-Domain Recommender Systems

“…Fewer works [8,16,53,69] studied the case of item overlap, and they all assume to have the same catalog of items across domains. Some works [2,9,23,35,60,61] studied the case of overlapping features, especially social tags. Shi et al [59] studied the sensitivity of the cross-domain recommender by varying the number of overlapping tags between 5 and 50.…”

A Novel Embedding Model for Relation Prediction in Recommendation Systems

“…According to the strategies of exploiting knowledge, the crossdomain recommendation approaches can be classified into two categories: 1) aggregating knowledge [24] and 2) linking and transferring knowledge [9], [27]. The approaches by aggregating knowledge from various source domain to perform recommendations in target domain include three cases: merging user preferences [21], mediating user modeling data [18], [24] and combining recommendations [10], and the models by linking and transferring knowledge between domains include three variants: linking domains [12], sharing latent features [13], [25] and transferring rating patterns [9], [20], [27]. Relation prediction.…”