Music IR: The State of the Art

• Retrieval

  • Gaps include polyphony, UI, indexing

• Evaluation

  • At least three projects have done
  • Two used Cranfield model, but relevance judgement/databases problems
  • No one has “real” relevance judgements or adequate databases
  • Need standard databases, “real” relevance judgements, more user studies
  • => Music equivalent of TREC

Previous slide

Next slide

Back to first slide

View graphic version