MUSIC IDENTIFICATION SYSTEM: ADDRESS METADATA ISSUES BY NOT FIXING IT

Slides:



Advertisements
Similar presentations
Rightsrouter and DOI Presentation to IDF Members Meeting 22nd June 2004 Gordon Rintoul - Director of Operations.
Advertisements

CONFIDENTIAL DIGITAL WATERMARKING ALLIANCE. CONFIDENTIAL DIGITAL WATERMARKING ALLIANCE 2 Digital Watermarking Alliance Charter The Digital Watermarking.
Publishing exhibitions using KE EMu 4 October 2005 Julian Tomlin Head of Administration.
Proprietary & Confidential The ScatterTunes Description.
Consider Acquisition of Patent Rights Explore A Mutually Beneficial Business Opportunity 1.
Who Controls The Internet?. Internet Radio and the DMCA.
THE PA or the “Composition” THE SR or the “Sound Recording”
Music Industry Quiz BTEC LEVEL 3 SUBSIDIARY DIPLOMA in MUSIC
1 ISAN International Agency 30 rue de Saint Jean CH-1203 Geneva Tel.: Fax: ISAN:
You manage, we deliver Gavin Robertson: VP Strategy and Biz Dev.
 Duplication (short-run) vs.  Replication (large volume-500+)  Packaging choices  UPC (Universal Product Code)  Selling CDs/distribution  Physical.
DELOVODNIK PRO A short presentation. Standard mail record keeping Most companies these days receive and send a lot of paper mail. Too many of them keep.
What types of resources will the collection contain? JPG images of The Cure’s album covers. Metadata about the items: format, identifiers, title, release.
How it Works - Seal-IT First Watermark Keycoded masterfile available DFile Identification and matching with UDM B (1) Owner is registered.
Shazam -Abdulshafil Ahmed -Steven Lewis -Rick Huang.
Garage Band Jingle Assignment MUM 2703 Professor Calle.
Achieving Semantic Interoperability at the World Bank Designing the Information Architecture and Programmatically Processing Information Denise Bedford.
Disc Jockey (DJ) From getting paid in drink tickets to millions of dollars a year. $300/night is normal.
+ The Learning Registry: A How To Primer for Digital Content Publishers and Aggregators December 20, 2011.
[1] National Institute of Science & Technology Technical Seminar Presentation 2004 Suresh Chandra Martha National Institute of Science & Technology Audio.
Open Source Security Updates Why it's Different; What you Should Know Josh Bressers Friday, 11 May 2007.
Professional Engineering Practice
Chapter 8 Sound FX Composition. Chapter 8 Sound FX Composition.
You integrate many IP from many suppliers?
16 Organizational Conflict, Politics, and Change.
Given by: Rose Rocchio, OIT, UCLA Nick Thompson, OID, UCLA
- A “Portable” Implementation
Section 9 Tagged File Support Design
Munix Overview Customer Profiling, Frequency Analysis, Location Mapping, Movement Analysis, VIP recognition, Time Analysis, and much more …. Internet Gateway.
Music Licensing 101 Webinar June 28, 2017 Click HERE for audio
Project Management: Messages
Record Production Chapter Thirteen
Pasquale Pagano CNR – ISTI (Pisa, Italy)
Data normalization. Integrity and Robustness.
A review of audio fingerprinting (Cano et al. 2005)
Chapter 18 Maintaining Information Systems
Creating Your Information Product
Music Licensing Chapter Seven
Copyright What we need to know. ©
UNDERSTANDING THE DIFFERENCE BETWEEN A SOUND RECORDING (SR)
Metadata Editor Introduction
Introduction to Music Information Retrieval (MIR)
Licensing.
MusicNet Announcement
Organizational Cost Reduction Steps to Use CRM Applications for.
An Overview of MPEG-21 Cory McKay.
Acquiring content legally and ethically
Choosing the Discovery Model Martin Forsberg
Best Practices: AP Automation for GP Dan Thomson
2e59c2a073542ecc6f cc8a99d0bb d34a369fddeef e9
SharePoint Essentials Toolkit
Chapter 7.3 The Publisher-Developer Relationship
denblogs.com/jendorman
Data Understanding, Cleaning, Transforming
Hands-on Introduction and Refresher Course
Ten Metadata Problems and Solutions
CTI STIX SC Monthly Meeting
Requirements Management
Mastering Audio 1.
The Basics of Information Systems
Institute of New Media Development and Research
Database System Architecture
PRODUCTION PHASES CHANGES
Subject Name: SOFTWARE ENGINEERING Subject Code:10IS51
The Basics of Information Systems
Console Game Development
FIJI Kamni Naidu and Shivendra Deo
BRIEF 4: Music Marketing/ Online
Continuing Education Solutions
Module 1.2 Value of the Master Facility List
Presentation transcript:

MUSIC IDENTIFICATION SYSTEM: ADDRESS METADATA ISSUES BY NOT FIXING IT Anoosha Chanda Senior Software Engineer @ Pandora

AGENDA Music Metadata Challenges with Music metadata Music Identification Technical Solutions Path to a cleaner catalog Agenda of the talk Start with an introduction to the challenges with music metadata, why it is broken and why we can’t fix it. Metadata role in music identification. Music identification systems and solutions to implement them. How to build a path to a cleaner catalog organization? What are the pressing problems that any music streaming service faces?

MUSIC METADATA Track, Album, Artist, Label, and Genre information Identifiers: International Standard Recording Code and Universal Product Code Post direct deals, deliveries from providers in DDEX format DDEX standard is a consortium of leading media companies, music licensing organizations, and digital service providers, focused on the creation of digital supply chain standards. ISRC – International standard recording code for a recording (Record label) UPC – Universal Product Code for an album ISWC – International standard work code (Publisher)

CHALLENGES WITH MUSIC METADATA Poor management of ISRC and UPC Record Labels and Artists agreements Multiple Deliveries from same Record Label Catalog growth Artists agreements with record label can change. The changes will result in multiple deliveries. Sometimes, conflicts can arise even though it was the same track we received earlier. Number of deliveries of a track can increase duplicates. It is not ideal to start fixing the existing content as the music content is increasing at a higher rate. Multiple Versions of Recordings

HOTLINE BLING MY WAY Hotline Bling by Drake (Single) Views by Drake (Original Album) R&B + Chill by Various Artists (Collaboration) Hotline Bling (Remixed + Reworked) by DJ Remix Factory (Remix) MY WAY Versions like live, remix, studio recording, remastered, credits etc. pose another challenge. The Main Event (K) by Frank Sinatra (Live) My Way by Frank Sinatra Tribute Band (Karaoke) My Way by Frank Sinatra (Remastered) My Way – Instrumental by The Gary Tesca Orchestra (Cover)

S T A N D B Y M E Want to take you all to a tour of Stand By Me for a minute. Reggae example

MUSIC IDENTIFICATION Primary usage of music metadata is music identification. We will discuss a couple of solutions to identify music without addressing the metadata inconsistencies directly.

AUDIO FINGERPRINTING eJzlnAmSJDeORa_EfTkO1_sfYR7g3cVSTgdTNaHsWkYyg8nCCRIAwY8PuqeMMdabi4jhJmq8iZkuwsZyE7XdxOwX4exVxHETc12EF59fi2JvYviLCPx7ESHeRLqKUW5i3.. 1. Each track is down sampled to 11kHz mono signal and passed through a whitening filter to reduce strong or stationary resonance in the signals that can be caused due to the type of recording. Then, we capture the relative timing between successive beat-like onsets detected in the audio. It captures the peaks in the spectrogram and stores them based on their time of occurrence as {hash, time} pairs, raw hash with time. 2. The raw hash code is stored on Solr. Solr is open source search platform. The code is encoded using Base64 encoder to create a fingerprint for easy handling.

PERFORMANCE MEASUREMENT Precision and Recall 𝑝𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛= 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 +𝑓𝑎𝑙𝑠𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 𝑟𝑒𝑐𝑎𝑙𝑙= 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 +𝑓𝑎𝑙𝑠𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒𝑠 In house technology vs. commercial solutions In short, we look for a system with not too many false positives (better precision) and not too many false negatives (better recall). In house technologies: you may use open source technologies to build it. OpenFP, Echoprint, Musicuri, audioscout Third party technologies: Gracenote, Audiblemagic, audioID, shazam and ACRCloud, a cloud platform that helps companies and developers integrate ACR Cons with fingerprinting system: even though it works well with most of genres, certain genres are not working as expected. For instance, classical, EDM Songs with different intro. Limitations in Classical, Reggae and EDM genres

GROUPING Recordings grouping: fingerprinting with metadata analysis Can also be applied to albums, artists, and labels etc. Open music encyclopedia Data science Grouping: fingerprinting and use other identifiers to generate metadata group. Automated grouping with ranking would be ideal metadata management system.

NEED FOR HUMAN CURATION Validation of Automated Systems Curation comes to rescue when automation fails. End with a quality control at scale: having curators, librarians and catalog managers to create a Segway to Taylor’s metadata. Fixing Clean vs. Explicit, Artwork, Sub titles etc. Quality Control at scale

PATH TO A CLEANER CATALOG Link International Standard Musical Work Code (ISWC) to ISRC Global music database ISWC - International Standard Musical Work Code - a unique, permanent and internationally recognized reference number for the identification of musical works. The ISWC identifies the musical work regardless of its copyright status, and regardless of agreements related to publication or the distribution of royalties in respect to the use of this work. This means that an ISWC will not change when a new agreement is made for the work, when the agreed distribution changes or when the work moves into the public domain.  ISRCs are usually provided by your label or distributor. ISWCs are assigned by “Performing rights organization” when your songs are registered. Each recording should only have one ISRC, but one song can have multiple ISRCs if the song has been recorded, remixed or edited more than once. Recording vs song identification number ISRCs to match master recordings to underlying compositions. ISWC helps to use your music distinguish your song from similarly titled songs and provides valuable metadata Linking ISWC to ISRC is challenging as we get ISRC from labels and ISWC for publishers. With the direct deals, we have access to ISRCs. Industry Projects to Link Compositions (ISWCs) and Recordings (ISRCs) DDEX launched a new Linking working group to tackle this issue. And, SoundExchange project was also started working towards addressing this issue.

THANK YOU