Show HN: 1M Song Dataset dev in 10 mins(mortardata.com) |
Show HN: 1M Song Dataset dev in 10 mins(mortardata.com) |
In the 6 minute example, they load a dataset from S3, then use Pig and Python to process it. You can "illustrate" each step of your code, which pulls out small, relevant samples from the dataset and shows the results.
I'm starting my thesis on music information retrieval, just studying the related work for now. If anybody has any suggestion on the directions I could follow would be really welcome.
My initial idea would be to focus on playlist generation taking into account user's history and usage. So far I've seen a lot of related work exploiting song similarity, some cool work on music mood and some on assisted playlist building. I'm also not ruling out recommendation or discovery.
I'm doing something similar for my master thesis, a pig console embedded in js and also Cassandra support. I expect to release it in mid-January.
I'm glad it looks awesome, thanks!
http://musicmachinery.com/2011/05/14/how-good-is-googles-ins...
There are some insightful comments from names I recognize from my research.