Conference Schedule

Many examples of large-scale music data mining use pre-extracted audio features in textual format as their input, such as the Million Song Dataset. But these datasets must originally have been somehow extracted from the raw audio, and this process needs to become more scalable to handle larger datasets more efficiently. This project uses Apache Hadoop to scale MARSYAS (Music Analysis, Retrieval and Synthesis for Audio Signals) into the cloud. Additional work has been done towards a recommender system, using Apache Mahout for scalable fuzzy k-means clustering of the extracted audio features.