• Douglas Eck (Google)
• Greg Mead (Musicmetric)
• Martin Roth (RjDj)
• Ricardo Tarrasch (Meemix)
• moderator: Rebecca Fiebrink (Princeton)

• rjdj – music making apps on devices like iphones
• musicmetric tracks 3 areas: Social networks, network analysis (influential fans), text via focused crawlers, p2p networks
• memix – music recommendation, artist radio, artist similarity, playlists.  Pandora-like human analysis on 150K songs – then they learn these tags with machine learning.  Look at which features best predict the tags.  Important question is ‘what is important for the listeners’.  Their aim is to find best parameters for taste prediction.
• google – goal is organize the world’s information.   Doug would like to see an open API for companies to collaborate

Rebecca is the moderator.

What do you think is the next big thing? How is tech going to change things in the near future?

• Doug (Google) thinks that ‘music recommendation is solved’ – he’s excited about the cellphone.  Also excited about programs like chuck to make it easier for people to create music (nice pandering to the moderator, doug!)
• Ricardo  (MeeMix) – the laid back position is the future – reach the specific taste of a user.  Personalized advertisements.
• Greg (MusicMetric) – Cloudbased services will help us understand what people want which will yield to playlisting, recommendation, novel players.
• Martin (RjDJ) – Thinks that the phone is really exciting – having all this power in the phone lets you do neat thing.  He’s excited about how people will be able to create music – using sensory inputs, ambient audio.

How will tech revolutionize music?

• Doug – being able to collaborate with Arcade Fire on online
• Martin – musically illiterate should be able to make music
• Ricardo – we can help new artists reach the right fans
• Greg – services for helping artists, merchandising, ticket sales etc.

What are the most interesting problems or technical questions?

• Greg – interested in understanding the behavior of the fans. Especially by those on P2P networks. Huge amount of geographic-specific listener data
• Ricardo – more research around taste and recommendation
• Doug – a rant – he had a paper rejected because the paper had something to do with music generation.
• Rebecca – has a MIR for music google group :MIR4Music
• Martin – engineering:increase performance in portable devices – research:how to extract music features from music cheaply
• Ricardo – drumming style is hard to extract – but actually not that important for taste prediction

How would you characterize the relationship between biz and academia

• Greg – there is lots of  ‘advanced research’ in academia, while in industry  there look at much more applied problems
• Doug – suggests that the leader of an academic lab is key to bridging the gap between biz and academia.  Grad students should be active in looking for the internships in industry to get a better understanding of what is needed in industry.  It is all about getting grad students jobs in industry.

Audience Q/A

• what tools can we create to help producers of music? – Answer: Youtube. Martin talks about understanding how people use music creation tools.   Doug: “Don’t build things that people don’t want.”  – to do this you need to try this on real data.

Hmmm … only one audience q/a.  sigh …

Good panel, lots of interesting ideas.  Here is the future of music:

A Multi-pass Algorithm for Accurate Audio-to-Score Alignment
Bernhard Niedermayer and Gerhard Widmer

ABSTRACT – Most current audio-to-score alignment algorithms work on the level of score time frames; i.e., they cannot differentiate between several notes occurring at the same discrete time within the score. This level of accuracy is sufficient for a variety of applications. However, for those that deal with, for example, musical expression analysis such micro timings might also be of interest. Therefore, we propose a method that estimates the onset times of individual notes in a post-processing step. Based on the initial alignment and a feature obtained by matrix factorization, those notes for which the confidence in the alignment is high are chosen as anchor notes. The remaining notes in between are revised, taking into account the additional information about these anchors and the temporal relations given by the score. We show that this method clearly outperforms a reference method that uses the same features but does not differenti- ate between anchor and non-anchor notes.

The main contribution is the introduction of an expectation strength function modeling the expected onset time of a note between two anchors. Although results are encouraging, there are specific circumstances where the algorithm fails, i.e., temporal displacement of notes is large.

A Cartesian Ensemble of Feature Subspace Classifiers for Music Categorization (pdf)

Thomas Lidy, Rudolf Mayer, Andreas Rauber, Pedro J. Ponce de León, Antonio Pertusa, and Jose Manuel Iñesta

Abstract: We present a cartesian ensemble classification system that is based on the principle of late fusion and feature sub- spaces. These feature subspaces describe different aspects of the same data set. The framework is built on the Weka machine learning toolkit and able to combine arbitrary fea- ture sets and learning schemes. In our scenario, we use it for the ensemble classification of multiple feature sets from the audio and symbolic domains. We present an extensive set of experiments in the context of music genre classifi- cation, based on numerous Music IR benchmark datasets, and evaluate a set of combination/voting rules. The results show that the approach is superior to the best choice of a single algorithm on a single feature set. Moreover, it also releases the user from making this choice explicitly.

An ensemble classification system built on top of Weka:

Results, using different datasets, classifiers and feature sets:

Execution times were about 10 seconds per song, so rather slow for large collections.

The ensemble approach delivered superior results through adding a reasonable amount of feature sets and classifiers.  However, they did not discover a combination rule that always outperforms all the others.

ON THE APPLICABILITY OF PEER-TO-PEER DATA IN MUSIC INFORMATION RETRIEVAL RESEARCH (pdf)
Noam Koenigstein, Yuval Shavitt, Ela Weinsberg, and Udi Weinsberg

abstract:Peer-to-Peer (p2p) networks are being increasingly adopted as an invaluable resource for various music information re- trieval (MIR) tasks, including music similarity, recommen- dation and trend prediction. However, these networks are usually extremely large and noisy, which raises doubts re- garding the ability to actually extract sufficiently accurate information.

This paper evaluates the applicability of using data orig- inating from p2p networks for MIR research, focusing on partial crawling, inherent noise and localization of songs and search queries. These aspects are quantified using songs collected from the Gnutella p2p network. We show that the power-law nature of the network makes it relatively easy to capture an accurate view of the main-streams using relatively little effort. However, some applications, like trend prediction, mandate collection of the data from the “long tail”, hence a much more exhaustive crawl is needed. Furthermore, we present techniques for overcoming noise originating from user generated content and for filtering non informative data, while minimizing information loss

Observation – CF systems tend to outperform content-based systems until you get in the long tail – so to improved CF systems, you need more long tail data.  This work explores how to get more long tail data by mining p2p networks.

P2P systems have some problems – privacy concerns, data collection is hard. High user churn, very noisy data, some users delete content from shared folders right away, sparsity

P2P mining Shared folders are useful for similarity, search queries are useful for trends.

Lots of p2p challenges and steps – getting IP addresses for p2p nodes, filtering out non-musical content, geo-identification, anonymization.

Dealing with sparsity:  1.2 million users, but average of 1 artist/song data point for each artist/song relation.  These graphs show song popularity in shared folders. They use this data to help filter out non-typical users.

Identifying songs: Use the hash file – but of course many songs have many different digital copies – so they also look at the (noisy) metadata.

Songs Discovery Rate

Once you reach about 1/3 of the network you’ve found most of the tracks if you use metadata for resolving.  If you use the hashes, you need to crawl 70% of the network.

Using shared folders for similarity

There’s a preferential attachment model for popular  songs

Conclusion: P2P data is good source of long tail data, but dealing with the noisy data is hard.  The p2p data is especially good for building similarity models localized to countries. A good talk with from someone with lots of experience with p2p stuff.