Digital asset management
Why metadata matters:
it greases digital wheels
Originally published in Current, Sept. 10, 2001
By Steven Vedro
If information isn’t tagged properly, in a way that speaks to the computers that oversee digital media, it might as well not be there. Public TV producers and everyone else in digital media will soon be working continually with metadata tags, so it’s worthwhile now to start preparing to use them.
Every video file shot by public TV producers, or edited or stored by master control, carries signature metadata tags that describe what the file contains, its format and its owner’s name. Such tags allow a music file to be played on the Internet. When we send out MPEG streams in digital television, the metadata PCIP codes will take care of listing programs on the viewers’ electronic program guide.
Indexing, storing, searching and retrieving material through metadata tags is at the heart of the emerging technology of digital asset management (DAM).
These tags make it possible for archivists and researchers to search video collections and for producers to create edit decision lists without ever touching the original video masters. They make video-on-demand possible and allow for secure e-commerce and licensing of video products. Metadata exchange is the technological underpinning of public TV’s proposed Online Education Service, the PBS traffic automation system and shared master control operations.
Indeed, these national efforts—and the program sharing that underlies all of broadcasting—won’t work without agreements about the metadata and its format. This article reviews the various international and industry-wide efforts now under way to agree on metadata terms. It aims to help stations begin thinking today about tagging their video products for the digital tomorrow.
"We are not unlike many public broadcasters in recognizing that the cost of holding, cataloging, archiving, and managing assets is becoming high," says Art Zygielbaum, director of research and development at Nebraska Educational Telecommunications. "The lost opportunity cost in not being able to identify and/or find reusable assets is also significant."
To index video, a producer must link metadata about the content to specific segments in an associated video file. "Ingest and indexing" systems from companies such as Virage, Convera and Mediasite take a video file and link video time-code to the text of the words spoken, closed captioned or superimposed on the screen. More sophisticated analysis tools—some not yet perfect—include speech-to-text, scene-change, and object- and face-recognition.
With information on program content, producer and so forth in an online database, a user can search for matching terms. The resulting list of "hits" can then be used (a) to create an online request for a hard copy of the video, (b) start a streaming video play of the requested "hit," or (c) download a digital video file.
DAM allows producers to rapidly preview indexed footage. Instead of watching video dubs over and over again to find possible edit points, a producer can search for specific clips based on the descriptors associated with the video files. The clips, accessible on the network used by everyone participating in production, are often lower-resolution "proxies" of the master files. Producers can use these proxies to create and preview edit decisions, with the DAM system aligning the resulting edit decision lists to the master file time codes.
The promotion staff can use these proxy versions to create text and graphics materials, and web designers can tap into this data to create interactive units based on the look and feel of the emerging video program. Production executives can track the use of the proxies, monitoring the approval process from idea to final cut.
DAM systems also help stations, producers and archive managers to find and deliver materials from their libraries online—to other stations, to institutional clients and customers, to researchers and to home viewers.
Producers can use footage already shot—an asset that otherwise would be lost—instead of shooting similar scenes anew. PBS Online’s partnership with Virage, for example, already allows viewers to search the closed caption text of the NewsHour and retrieve short segments that contain the entered search terms.
Contextual metadata adds value
The real power of DAM comes from knowing the words spoken and the images captured inside a video clip and linking them with information about them: what a clip is about, who owns it, who can use it, and so on. This information does not create itself—people have to apply agreed-upon classifications.
These metadata clearly are essential in education. By adding additional contextual keywords such as grade level, curriculum goals and standards, an indexed DAM database can serve as the portal to videos that teach specific facts and concepts. Some of the earliest work in this area has been done by Kansas City’s KCPT and South Carolina ETV.
Chalkwaves Digital TV Handbook (www.chalkwaves.org) was created by KCPT in cooperation with WSIU (Carbondale, Ill.), Smoky Hills PTV (Bunker Hill, Kan.), and KPTS (Wichita, Kan.). Users can search its database for keywords in more than 1,000 educational program segments in the stations’ combined libraries. Each program has its own web page, which gives detailed information about it and an order form for the program’s teachers guide. The page also offers downloadable lesson plans created specifically for each series and links to vetted sites that correlate with the video, including streaming video professional development "E-Tools." Chalkwaves provides a two-minute streaming video preview clip for each program. The program itself is delivered via videocassette to the recipient’s school library.
South Carolina ETV’ s SearchSCETV (cfmedia.scetv.org/searchlite) extends the ITV catalog to a more detailed level. With funding from the governor’s K-12 Technology Initiative, the state network digitized more than 800 hours of its library of locally produced programs. Ingest software identified scene changes that mark the start points of "content idea" segments of three to five minutes in length. A cataloging staff entered keywords and appropriate Decimal Classification numbers and Sears Subject headings for each segment into an Informix database.
In March , South Carolina ETV incorporated SearchSCETV into a larger online reference service, www.knowitall.org, which also includes digitized and key-word searchable segments from its NatureSCene video series and website, "virtual reality" tours of historic locations and a number of formerly inaccessible still-image collections from the state library. The "backpack" feature allows registered users to download clips they have previewed. The system minimizes copyright concerns by permitting downloads only to Internet addresses in South Carolina.
Moving to large-scale DAM
While DAM is fairly new to the broadcast industry, it has been the subject of much research in the university community. With large library systems and media collections and thousands of users, universities have installed a number of large DAM systems.
Vendors include IBM (Content Manager, www-4.ibm.com/software/data/cm/), Ascential Software (Informix Media360, recently purchased by IBM, www4.ibm.com/software/data/informix/news/), Artesia (TEAMS, www.artesia.com) and Bulldog (www.bulldog.com/view.cfm).
Front-end video ingest systems bring data into much larger database systems that manage diverse forms of media and track detailed information about each indexed asset. Information can include ancillary documents such as interview notes, music clearances and rights information. The systems automatically deliver digital proxies in multiple formats to multiple platforms such as broadcast servers, web pages and physical media. They can be linked to digital rights management (DRM) systems and various e-commerce products.
Given their complexity and cost, it is no surprise that the early DAM efforts within public TV are coming from major production centers such as WGBH and WNET, big university media centers in Nebraska and Utah and national services such as the BBC and the Mexican Ministry of Education.
A shared system in Utah
The University of Utah’s CAMS (Content Asset Management System) project is being coordinated by KUED Media Solutions and its partners KUED Television and the Utah Education Network (UEN, the state’s ITV and instructional media agency). CAMS will support all three organizations as they create and share libraries of learning materials for classroom use and web delivery and build video-on-demand and multicast streams for KUED and UEN.
The project was prompted by the organizations’ "need to re-purpose and recycle the media assets we have, increase our efficiency in managing media resources and allow our various clients to share in the intellectual wealth," according to Paul Burrows, self-described "Yenta for Technology and Media" at KUED Media Solutions. "A DAM system works for an organization by helping it grow new assets, wrangle existing assets and then drive them to market," says Burrows.
The project chose IBM’s Media360 system "because of its scalability, enterprise-wide management and distribution potential," Burrows said. Virage VideoLogger will serve as the primary video ingest tool.
The initial rollout of the service will focus on a "realistic workout" of the integration of video and speech-to-text ingestion and encoding, workflow and metadata indexing, as well as search-and-browse functions.
Indexing WGBH’s treasures
In Boston, WGBH’s TEAMS Project is building a DAM system to organize and index more than 160,000 hours of archived content contained on more than 311,000 tapes. The station hopes that carefully indexed content will increase the business of its "footage sales" units—both for internal and external customers.
The big producing station is working with Artesia’s TEAMS software developers and metadata designers to build a workflow system to identify, track and control access to production-related video, audio and text documents, including rights management data. The system protects the integrity of the master file while selectively permitting edits of multiple versions of program segments as they move among producers, editors, web designers and promotion staffers.
Coordinating Mexico’s distance learning
Before implementing its DAM service (using Media360 software), Mexico’s Instituto Latinoamericano de la Comunicacion Educativa (ILCE, www.ilce.edu.mx) had more than 100,000 hours of video programs sitting on shelves. When a teacher needed to deliver material to a class on a subject such as disappearing species, the instructor could query ILCE’s website. The query would produce a list of pertinent videos. The teacher could select one or more for transmission and set a VCR to record the subsequent broadcast.
Media360 is enabling ILCE to transform this process. Key programs are ingested using the Virage Video and Audio Loggers, with metadata stored automatically in Media360. When a teacher asks the ILCE website for videos on "endangered or disappearing species," she now sees complete transcripts (created by voice-to-text conversion of the original material), keyframe images and related time-codes. The instructor can browse the keyframes or text, selecting the most appropriate sections of the program for class use. These segments can then be broadcast, sent by FTP to the school computer or delivered on a CD-ROM. If the school has a fast enough Internet connection, the teacher can preview programs.
Standards: the labor-intensive part
DAM systems provide sophisticated tools to describe all forms of media assets, from whole programs to the smallest "grain" of content. Once labeled, content can be found, routed to the correct destination and played as many times as rights permit, leaving a digital trail for payment and security purposes.
What these systems can’t do by themselves is ensure that these descriptors mean the same thing to all parties—to producers and distributors, between competitors and across industries.
- Master-control and traffic systems need to know what programs they are receiving.
- Hardware devices need to know what formats need to be decoded.
- Archivists need to know how to classify what they put on the "electronic shelf."
- Educators want to know the instructional attributes of the materials they find in database searches
Advocates for every need are involved in standards-setting efforts. Without such standards, metadata become a meaningless jumble.
Because metadata are so critical to all of these automation plans, a number of heavyweight, world-scale organizations are working on different aspects of DAM taxonomies.
SMPTE—the Society of Motion Picture and Television Engineers—recently published a recommendation for a Unique Material Identifier and a Unique Program Identifier, and is developing a "metadata dictionary" for the video industry and the MPEG-7 standard.
The BBC has developed the Standard Media Exchange Framework (SMEF), a set of references for indexing and archiving video production content. One aspect of this is the SMEF Data Model (SMEF-DM) consisting of a set of metadata definitions for the information required in production, distribution and management of media assets, currently expressed as a data dictionary and set of Entity Relationship Diagrams.
In the educational community, a number of projects are creating classification schemes for "learning objects," including video and multimedia materials. The International Standards Organization (ISO) has created a joint technical subcommittee (SC36) on learning objects metadata (jtc1sc36.org) charged with bringing together the two major media classification systems used in education: Dublin-Core (www.sics.se/~preben/DC/DC_guide.html) and IMS (Instructional Management Systems, www.imsproject.org/metadata/mdinfov1p1.html).
Under the Institute of Electrical and Electronics Engineers (IEEE), a Learning Technology Standardization Committee (ltsc.ieee.org/doc/wg12/LOM_WD4.htm) has created a Learning Object Metadata group (LOM). This group has defined learning objects as "any entity, digital or non-digital, which can be used, re-used or referenced during technology-supported learning." Most early discussions about learning objects assume they will include video and other media elements supporting a 3-10 minute learning activity.
The U.S. Department of Defense has requested that all providers of instructional content to the armed forces create a common set of metadata standards called Sharable Courseware Objects Reference Model (SCORM, xml.coverpages.org/scorm.html), based on the XML technology now being introduced on the Web. Using SCORM codes, military educators will be able to exchange, manage, track and reuse learning content, regardless of source or use. Producers of government training videos will soon need to meet these classification requirements. The Advanced Distributed Learning "Co-Lab" at the University of Wisconsin (www.wiadlcolab.org) will consider how to extend SCORM classifications to instructional video objects.
What metadata will public TV need?
While these various international bodies are at work creating the high-level dictionaries for classifying all media content, leading-edge public broadcasters are not in a position to just sit still and wait. They are launching DAM systems knowing that they will want to leave room for additional fields and compatibility with international standards that are not yet fully defined. At KUED Media Services, the CAMS team has spent hundreds of hours designing their workflow and reviewing and adjusting the "out-of-the-box" metadata taxonomies provided by Media360.
In Utah’s system, more than 65 fields will identify the "owner" and the "source" of the indexed object. Even more fields will be devoted to describing the object’s external attributes, such as program title, series, media format and archive location. The system developers in Utah are making room for other content attributes such as detailed subject fields, place and activity represented, people involved and their affiliations, using Dublin-Core and IMS educational content indexing templates.
For example, a few of the many elements in the IMS educational template are: interactivity type (active versus expositive), learning resource type (exercise, simulation, questionnaire, and so on), intended user (teacher, author, learner, manager) and difficulty level (from very low to very high).
At WGBH, with its vast archive of content and scores of different applications and production teams, the station will use hundreds of indexing fields. The first step, according to David Liroff, v.p. and chief technology officer, is to begin the internal process of tracking workflow and of capturing all of the ways content is described. This is a long and slow process, says Liroff. "With over a decade of experience in building video content databases, we are still discovering that different departments have different terms for the same data, or use different data elements to represent the same term," he says.
PBS, CPB and OES initiatives
While PBS has delayed putting out a formal request for proposals for a national DTV-Advanced Traffic Software system, the network still intends to pursue DAM technology for public TV, according to John Tollefson, chief technology officer. Indeed, he says, the top four PBS technology projects involve metadata and must be coordinated with each other and with DAM activities at stations.
Meanwhile, CPB has begun a metadata coordination effort led by Alison White of the Television Future Fund. "It’s critical that stations and national organizations work in concert to develop shared metadata standards to thread through all of our work, from inception to delivery," says White. One of the first studies will look at stations’ DAM needs and activities.
Art Zygielbaum of Nebraska ETV, a participant in CPB’s March 26 planning meeting on DAM, says PBS and CPB should survey metadata efforts by stations and "glean commonality" that could lead to system standards. David Liroff agrees. It’s "comforting" to him, he says, to see PBS and CPB taking leadership roles in recognizing the interrelationships between DAM and so many emerging technology projects.
Another CPB-initiated project may also emerge as a catalyst to move public TV and instructional video producers to a common metadata standard. The proposed Online Educational Service (OES), as proprietor of a national educational content database, will need to ensure the accuracy and consistency of metadata maintained at dozens of provider nodes and organize a common set of search terms. It will coordinate multiple servers to send users the appropriate proxy in the requested format, track its delivery and ensure its appropriate use.
While it’s unlikely that OES will specify a specific DAM software product to its content partners, it will have to establish procedures for sharing metadata terminology among content providers, explains Doug Weiss, CPB’s v.p. for strategic development.
"One of the compelling rationales for the development of OES is the potential for stations to leverage educational content resources across the public broadcast community," says Weiss. "The vast majority of such resources were originally archived for the purpose of facilitating on-air broadcast operations. It’s not too difficult to determine a title, date of air, location or other basic information about what is stored in station vaults. A more precise catalog indicating, let’s say, what curriculum-related subjects or concepts are covered in a particular program simply does not exist. It will have to be created."
Weiss, too, is looking for shared definitions that may likely come from outside of the broadcast industry. "While it might be tempting to simply go off and invent a system of classification," he says, "we are reminded that in this brave new media world, broadcasters—both public and commercial—are not leading the charge. Emerging media classification standards with bewildering acronyms are fast becoming de facto arbiters of exchange."
Come on in, the water’s wet
While you do have to swim through a sea of bewildering acronyms to reach any understanding of DAM metadata issues, there are important reasons why station leaders should at least get their feet wet today. Any organization producing content needs to start establishing immediately a basic process for building databases so that its content is not lost to future producers, web managers and clients.
As David Liroff told the heads of the national public broadcasting organizations at a recent dinner, "every public television station is evolving into a digital library—a community-based point of access for on-demand educational and public service media and related services.
"As public broadcasters," Liroff says, "we have an obligation to realize maximum value from the editorial content we create and acquire. It is our principal asset, the dowry we carry into the future." Liroff cites Metcalf’s Law, arguing that the value of public TV’s aggregated content will increase by the square of the number of stations that share the same coding schemes and practices.
This effort is not something just for the big guys. "Every station carries, or will carry, closed captions, descriptive video, PSIP for program guides and other program-related data, and other data to enhance the viewing experience," says John Tollefson. "Every station has an interest in being able to access programs and data about the programs, so that public television develops databases, procedures, and other infrastructure to serve the entire American public. Our educational mission drives us to look for methods to supply an enriching experience to individuals and educational institutions. DTV, for the first time in our history, provides the technical means to fulfill this mission. The challenge is now ours to use the technology wisely."
Thinking about "data about data" is part of that challenge.
Steven R. Vedro is a technology journalist, facilitator and consultant based in Madison, Wis. He has been involved in metadata planning as a technology consultant to CPB and the Online Education Service and is helping plan a session on the subject in Madison on Sept. 10-11 for the Reforging Links Project, which he coordinates.
Is that learning object the right one for me?
The electronic engineering profession is already mapping out categories of metadata that will describe digital learning materials for educators. These features come from the IEEE Learning Technology Standards Committee’s Learning Object Metadata Version 3.5:
interactivitytype (the type of interactivity supported by the learning object)
learningresourcetype (specific kind of resource, most dominant kind first)
interactivitylevel (level of interactivity between an end user and the learning object)
semanticdensity (subjective measure of the learning object’s usefulness as compared to its size or duration)
intendedenduserrole (normal user of the learning object, most dominant first)
context (the typical learning environment where use of learning object is intended to take place)
typicalagerange (age of the typical intended user)
difficulty (how hard it is to work through the learning object for the typical target audience)
typicallearningtime (approximate or typical time it takes to work with the resource)
description (comments on how the learning object is to be used)
language (user’s natural language)
To Current's home page Related articles: Current Briefing on digital TV. Outside link: CPB's website for stations includes a section on Digital Asset Management.
Web page posted Dec. 3, 2001
The newspaper about public television and radio
in the United States
A service of Current Publishing Committee, Takoma Park, Md.