WhatBy Edward Cone | Posted 2003-02-01 Email Print
The media giant's publishing arm broke all its books into thousands of digital assets. But was it too late to win over the rest of the company?s An Asset, And Where Are They?">
For instance, before the Book Group could begin distributing its assets, it first had to determine (a) what one was, and (b) how it would find each, once stored in an archive.
The project team in early 2000 included Gore, the technology project manager for the digital asset system; Madans, who ran the project from the business side; other Book Group staff; representatives from software vendor Artesia (see Dossier); and project managers from Accenture. The team met often with groups of users from 15 departments across the company.
"We went through the whole process of what files they had, how they got them, who controls that," Madans says. "We did workflow charts for every department, identified what assets they had in common."
The team spent about three weeks defining what an asset was. Changeable pieces of information, such as price lists, delivery dates or production schedules, were deemed not to be assets, while sample chapters and jacket copy made the original list of 33 asset types. Some were obvious, like covers and photos, while others took some pondering, like audio clips.
The key word was "reuse," Madans says. "It had to have a purpose to be in there, so that one department could pick up an asset in finished form and use it, or create a new asset out of it." A description of a book, for instance, could be reused as copy for a jacket flap, promo copy for an online or print ad, or summaries for online booksellers.
In the case of Four Blind Mice, the entire text is one digital asset. But a file of sample chapters is a separate asset that hooks readers at JamesPatterson.com. By the time Four Blind Mice hit stores in November, its cover, author bio, review blurbs and other information had been pushed out to sales reps and customers many times and for many uses, from sales presentations to online display. Ingram, the big Nashville-based wholesaler and jobber, had redistributed the same material to the chains and smaller bookstores it serves.
Now, Leung says the Book Group is considering allowing major accounts to pull some information directly from the book division's repository of digital assets, to speed up the process and take the task of chasing down assets off her hands.
But just because a discrete piece of a book can be identified doesn't mean it becomes a separate digital asset. The title, for example, describes the book and is thus considered not an asset but metadata. Metadata is descriptive information about an asset that helps it get found. "Production schedules are not an asset, but things they track, like first-passed pages, are assets," Madans says. "We had to do a lot of brainstorming" to draw the line.
The trick when assigning metadatawhich is really data about datais to use as many descriptors as needed to cover the logical means of identifying a digital asset. But not so much that retrieval becomes unwieldy.
Descriptive information for the Book Group's system includes the publishing industry identification code, known as the ISBN, as well as author, date of publication, genre (e.g., fiction), format (hardback or soft cover) and so on. Some metadata, like the Book Industry Study Group standard used by retailers to place books, was available from the publisher's existing databases.
A wall-sized chart of asset types and metadata was assembled in a crowded conference room that served as project headquarters on the 45th floor of the Time-Life Building. "It was a matrix with asset types across the top and metadata down the side, and in between was how that metadata interacted with that asset type," says Madans.
Thus the top row of the metadata/asset matrix would include asset types such as "fact sheet" or "production manuscript." Below an asset type would be a description of the formats required for that asset, whether Word or PDF for a manuscript or Quark for a jacket mechanical. Under that went information on who was responsible for importing the asset to the system; for example, the art department putting in a cover image. And on the left-hand side were the metadata elements needed for each asset type. This was precision work. Some metadata pertains only to one particular asset type, such as the "foil status" of covers that may be printed on foil-like material.
Creating metadata can be even harder for other media, such as video. Moving images don't come with captions, much less headlines, bylines, summaries and descriptions. "It's really hard to describe some unstructured information," says IBM's Brett MacIntyre. "A name has maybe 25 characters, but a two-hour video has all those imageshow do you describe things when you are dealing not with measurements but attributes?"
So a clip of a Kodiak bear eating a salmon in Alaska might be tagged with key words like, "Alaska," "salmon," "bear," and "eating." More key words might be added to images from key scenes.
In effect, some human being has to translate visual information into descriptions. At AOL Time Warner's CNN unit, video in the digital-asset management system is described in three different ways: by standard text labels, by digital transcripts of the audio track and by the images at each point where a scene changes. Still it's hard sometimes to determine when a video sequence starts and endswhether "bear eating" includes the bear catching a fish or just chewing on itmuch less describe what's in it comprehensively yet concisely.