Making Metadata: The Case of MusicBrainz

Citation: Jess Hemerly (2011/05/05) Making Metadata: The Case of MusicBrainz.
DOI (original publisher): 10.2139/ssrn.1982823
Semantic Scholar (metadata): 10.2139/ssrn.1982823
Sci-Hub (fulltext): 10.2139/ssrn.1982823
Internet Archive Scholar (search for fulltext): Making Metadata: The Case of MusicBrainz
Download: http://www.vigliensoni.com/McGill/CURSOS/2012 01/MUMT609/9 REFERENCES/making11hemerly.pdfd/making11hemerly.pdf
Tagged: musicbrainz (RSS)

Summary

"study employed both quantitative and qualitative research methods, beginning with a survey administered to the MusicBrainz community and data scraped from user profiles, followed by observation and qualitative interviews with registered users, called editors. Qualitative and quantitative data were analyzed and interpreted at the same time. The study sought to answer the following research questions about MusicBrainz:

• Contribution: Why do people contribute? Who are the MusicBrainz contributors? What characterizes editors’ participation? How is contribution linked to cultural preference? How can one compare contributors’ motivations in MusicBrainz to other constructed cultural commons?
• Music Information: How does MusicBrainz negotiate guidelines and standards? What is the relationship of MusicBrainz to other music resources? What is the role of metadata in music technology?

Findings are split into four categories: Demographics, including information about age, gender, region, and overall editing statistics; How MusicBrainz Works, an overview of the technical components and governance structure; Patterns & Processes, which describes the links between musical taste and contribution, how MusicBrainz serves as a tool for discovery, and the ways editors’ decisions mimic those made by information professionals; and Attitude & Motivation, which examines intrinsic and extrinsic motivations that drive users to contribute, from their belief in the philosophy of open source to a compulsion for accuracy and consistency."

Describes what metadata is, basic application and challenges for music. Notes NISO definitions of descriptive, structural, and administrative metadata as not always mutually exclusive applied to music.

Notes neither major commercial music metadata providers (Gracenote and AllMusicGuide) are peer produced, peer edited, or truly free.

MusicBrainz founded after Gracenote privatized CDDB.

"Catherine Marshall and Frank Shipman ... discuss the tradeoffs involved when using amateurs to reduce metadata costs and the importance of trust in the authority of a metadata source. With MusicBrainz, the negotiation of style guidelines and the metadata’s entry into the database are all tasks performed by a community of non-professionals. However, these non-professionals are dedicated to high quality, and, as this paper will explore in great detail, have developed standards and authority control to achieve that level of quality. The data is of high enough quality that, in 2007, Last.fm and BBC Music reached commercial licensing agreements with MusicBrainz, and the project has provided metadata to both sites since."

Briefly describes commons, states "MusicBrainz is a constructed cultural commons for music metadata." MusicBrainz potentially faces free riding problem. Briefly describes contributor motivations for Wikipedia and FLOSS.

Describes survey, MusicBrainz user profile scraping, and interviews.

Survey respondents overwhelmingly male (241 of 248) and in Europe (59%) or North America (31%). From profile scraping, describes contributors by longevity and number of contributions.

Describes basic interaction with MusicBrainz via music tagging applications, website, and communication within community.

"Representation of the data, down to hyphenation and capitalization, is governed by community-established guidelines. The guidelines are intended to be flexible suggestions to guide users in things like capitalization, punctuation, and representation, but editors take them quite seriously and many feel that the guidelines are strict rules only to be bent in special cases. As a result, the guidelines are more like community-established standards than flexible suggestions. Editors who violate the guidelines without having proper evidence that an artist intended it as such face the possibility of having their edits rejected by the community.

An edit is committed to the database as soon as it receives three unanimous “Yes” votes, and fails when it receives three unanimous “No” votes. Two “Yes” votes and a “No” vote will hold the edit in the queue. Editors can also choose to “Abstain,” which is a public way of declaring, “I don’t know.” If an edit of default quality receives zero votes, the edit will be automatically accepted after 14 days (the majority of the data in MusicBrainz is “default quality”). Editors are not able to vote on other edits until they have had 11 edits accepted and have been members for at least two weeks, with a confirmed email address. The voting mechanism gives editors time to review changes to the database and more quickly approve those things which are deemed correct and reject those things which are incorrect or do not conform to the style guidelines.

Debate about major changes to the style guidelines occurs via the Style Council mailing list. Any editor can be a member of the Style Council by simply signing up for the mailing list and joining the conversation. An editor proposes a change by sending out a request for comments (RFC) to the mailing list, to which other “council members” then provide comments and feedback. As one explained, if the comments are mostly negative, the submitter must either make suggested changes or withdraw the RFC altogether. If the comments are mostly positive, , then he RFC moves onto the next stage, which is Request for Veto (RFV). Everyone should have made any comments they were going to make on the RFC, but at the RFV stage, any Council member can also veto the change. If a council member vetoes it, the proposal goes back to “square one,” as one puts it [20]. If nobody vetoes, the change is approved."

Adding new releases most common contribution, linked to acquiring new music. Adding information deemed by contributors easier than voting on others' additions. As of January 15, 2011, 97.4% of edits were live, 1.3% had been deleted, and another 1.3% had been voted down. Votes on edits broke down as 55.8% yes, 39.5% abstain, 4.6% no.

Discusses contributor motivations and taste, allegiance to open source/commons, enjoyment, obsessiveness. Contributors who discovered MusicBrainz through Last.fm make more contributions.

Future Work:

• Comparative analysis of Wikipedia and MusicBrainz
• Comparative analysis of FLOSS and MusicBrainz
• Comparative analysis of other music info sites and MusicBrainz
• Women and commons-based peer production
• MusicBrainz and the future of music metadata
• Apply Madison et al's framework for studying constructed commons to MusicBrainz (in progress; draft)
• Mapping behavior of users of both Last.fm and MusicBrainz

Theoretical and Practical Relevance

Author interview with Rob Kaye, MusicBrainz founder, and class project page.