作者: Claus Aage Jensen , Ester M. Mungure , Torben Bach Pedersen , Kenneth Sorensen
DOI: 10.1109/ICDEW.2007.4400975
关键词: Computer science 、 Data modeling 、 Digital audio 、 Similarity measure 、 Data model 、 Metadata 、 Information retrieval 、 Music information retrieval 、 Similarity (geometry) 、 Identity (object-oriented programming)
摘要: Motivated by the increasing amount of digital music on WWW, this paper presents a data and query model for dynamic playlist generation. Queries are continuous, i.e., songs retrieved one at time, allowing user to dynamically influence retrieval. The can support arbitrary similarity measures, requiring only identity properly metric space be obeyed. Using measure, we able retrieve similar given seed song avoid retrieval disliked songs. Additionally, allows combined querying both metadata musical content. Vie has been implemented in prototype system that is very large number simultaneous users.