Google Books: A Metadata Train Wreck?

A recent post on Language Log titled “Google Books: A Metadata Train Wreck” by Geoff Nunberg argues that the errors in Google Books’s metadata are egregious and unforgivable. He wrote the blog entry after presenting his conclusions at the Google Books Settlement Conference. He cites many examples of poor metadata in Google Book’s collection – dates that are so wrong they’re funny, searches that return impossible results, and ridiculous classification errors. Nunburg says that Google should take responsibility for fixing the metadata problems, and he questions whether Google’s engineers should be trusted with the “Last Library,” as he calls it.

Nunberg reports that Dan Clancy of Google responds to his charges by saying the errors are the fault of the libraries (and metadata providers) supplying Google with the content. According to Nunberg, Dan Clancy also suggests that users could fix the errors one by one, in the way they fix errors on Wikipedia.

Finally, Nunberg suggests that the Antitrust Division of the Justice Department should consider Google’s poor metadata design and implementation resulting from “no contractual obligation, and only limited commercial incentives, to get it right.” Check out the comments to the post - they're pretty interesting as well.

We’re all probably familiar with the quality vs. quantity question: do you use your resources to digitize 1,000 items with great metadata or do you use your resources to digitize 1,000,000 with not-so-good metadata? Crane touches on this when he talks about noise. Lynch seems to argue that large collections – even with not-so-good metadata – have the potential to be more useful because they provide a larger “data set” on which to aggregate and recombine information. Are Google’s errors unforgivable, or are they an unfortunate but inevitable consequence of creating such a large collection? Who should take responsibility for them? Is Google the “curator” of Google Books?