BBH Central IconBBH Report Home PageSandy and Dave
  CENTRAL home  |   REPORT home About/Contact Us  |   Subscribe  |   Index by Topic  
The June 22, 2004 Issue Provided by System Dynamics Inc.
Table of Contents Print this article Email this article to a friend

More on Metadata -- A Dialog with Gracenote

Several months ago, we wrote about our experience with GetDigital -- see Get Digital Part 2: It's The Metadata, Stupid!. We compared GetDigital with CDDB, the online database used by most PC programs to obtain metadata when playing CDs and recording or "ripping" them to hard drives.

We first wrote about Gracenote more than two years ago -- see CDDB - Gracenote. At that time, their CDDB database was maintained almost entirely by end users, leading to inconsistency in data entry and classification.

Shortly after we published the GetDigital article, we received a nice note from Kathryn Shantz of Gracenote, pointing out CDDB improvements that we were not aware of. That led to an exchange of email with Kathryn and a telephone interview with Ross Blanchard (Director of Marketing and Business Development).

CDDB

In her email, Kathryn said "it's great that you're writing about metadata - while relatively hidden, this is one of the most important aspects of the digital music revolution! Gracenote's business was built around serving this need -- with the goal of making digital music easier to use." She described "some of the things we've changed to improve and extend the service since you last wrote about us:

  • In-house team of database editors who add new content and manually review and approve incoming content
  • International editors who check and add content in many different languages
  • Over 15,000 automated rules, filters & heuristics check over 20 data fields for consistency, accuracy and completeness
  • Improved genre categories so that it's easier for people to find exactly the right music for that moment, and editorial oversight of genre classifications"

In our telephone discussion with Kathryn and Ross, we learned that the CDDB database now includes nearly 3 million CDs and more than 36 million songs. Each day, users from ninety countries submit about ten thousand new albums to CDDB. In the US, Korea and Japan, Gracenote has editorial staff to "vet" the entries and "lock them down" to prevent modification by users. Because of the volume of entries, the editors are focused on the most popular content; of the most popular CDs, "the high 90% are locked down."

For the smaller labels, Gracenote has a "content partner program" so that they can directly submit album and track metadata for their own CDs and lock them down.

Gracenote does not lock down the genre metadata, which is highly subjective. They allow for two different genres for each album, artist and song.

Access to the CDDB catalog varies a great deal by country. For example, in Germany, access is 30% local (German CDs) and 70% international, while in Japan and Korea it's 70% local and 30% international.

Gracenote licenses a CDDB Software Development Kit (SDK) to software developers. Most PC-based CD players and recorders use this kit, which is also available for Apple and Linux-based systems. Gracenote also provides an "Embedded CDDB" solution; for portable devices and others without an active Internet connection, this provides a copy of CDDB for storage on a hard drive.

MusicID

Gracenote has a new system called MusicID which uses waveform analysis to create a "fingerprint" of each track; the track can come from any source: records, cassettes, CDs or even radio. (This is very different from the original CDDB recognition, which is based on track-length information specific to CDs.) Once the track is identified, the CDDB database provides all the metadata. Gracenote now has waveform fingerprints for over 3.6 million tracks, and says it is adding "over 25,000 a week."

Gracenote has patents pending on the recognition technique, and offers a MusicID SDK for license to device developers. We asked whether any products were available yet, and Ross said "it has been licensed to half a dozen folks - there's an Onkyo product but it's for the Japanese market only."

We have a large collection of LP records, and we've started recording them to a hard drive so we can play them through our network. We've been doing it with what's available, and it's a pretty tedious process; since nothing we've seen includes a link to CDDB, we have to enter the metadata ourselves. We're hoping to see someone bring out an LP recording system using MusicID to make it a lot easier; maybe it's not so far away.

( www.get-digital.net ) ( www.gracenote.com )