Improving Auto-tagging by Modeling Semantic Co-occurrences

@inproceedings{Miotto2010ImprovingAB,
  title={Improving Auto-tagging by Modeling Semantic Co-occurrences},
  author={Riccardo Miotto and Luke Barrington and Gert R. G. Lanckriet},
  booktitle={ISMIR},
  year={2010}
}
Automatic taggers describe music in terms of a multinomial distribution over relevant semantic concepts. This paper presents a framework for improving automatic tagging of music content by modeling contextual relationships between these semantic concepts. The framework extends existing auto-tagging methods by adding a Dirichlet mixture to model the contextual co-occurrences between semantic multinomials. Experimental results show that adding context improves automatic annotation and retrieval… CONTINUE READING
Highly Cited
This paper has 23 citations. REVIEW CITATIONS