Fixing Transcripts with the Crowd
A crowdsourcing project promises to demonstrate that when it comes to providing access to audiovisual archives, not only users with visual or hearing disabilities benefit. All users may.
Universities and colleges, along with other varieties of institution, are finding that, if they are to advance their businesses by providing public media content, then transcriptions and other accommodations of people with disabilities are legally required. The Americans with Disabilities Act of 1990, like similar legislation in other countries, demands it.
But advocates of access for people with disabilities have long noted that many kinds of users benefit from improved access. A project of WGBH, the large Boston public-television station, may serve as another of many illustrations of that.
To complete and correct transcriptions for some 40,000 hours of television programming held by the American Archive of Public Broadcasting, its collaboration since 2013 with the U.S. Library of Congress, WGBH has just launched FIX IT, an online game in which supporters of the station and the Archive can fix the many errors that result when producers rely on machine-generated transcriptions of audiovisual media, at least for a first approximation of transcriptions.
In the game, supporter-players view short audio clips drawn from video recordings and identify and correct errors in accompanying fragments of caption. That will permit AAPB archivists to create accessible and searchable transcripts and indexes for the Archive’s digitized public-media content.
Automatic transcription is infamously imprecise; for all their sophistication, voice-recognition algorithms have a ways to go, if indeed they even have a plausible destination. That hasn’t stopped automatic transcription from being rolled out on many platforms, including YouTube. WGBH is among several non-profit organizations that have worked with Pop Up Archive, a platform of tools for organizing and searching digital spoken word. It is also working with the University of Texas at Austin’s HipSTAS, a project that is creating an archive of speaker voices and associated sound characteristics through soundwave analysis and machine learning.
In theory, such technologies hold great promise: you can imagine the day when speech can be instantly converted into text, so that people with hearing disabilities can follow news reports, movies, and much else, and any user can have access to transcriptions of audiovisual material very soon after initial broadcast.
For the moment, that remains an elusive goal, so media content providers, such as television producers, must rely on human input, which is costly and slow. FIX IT will conceivably greatly accelerate the process.
Not only people with, for example, hearing or reading disabilities would then be able to use the archived material more fully. So would journalists, historians, and members of the general public for whom the archives suddenly would become far more searchable, and accessible.
WGBH and AAPB received funding assistance for the project from the Institute of Museum and Library Services, which provided a grant in 2015 so WGBH could explore “enhanced methods of providing access to audio-visual collections.” The Archive is a national effort to preserve at-risk public media before its content is lost, and to provide a central web portal for access to 60 years of public-station programming.
“For grammar nerds, history enthusiasts, and public media fans,” WGBH said in announcing the project, “FIX IT unveils the depth of historic events recorded by public media stations across the country and allows anyone and everyone to join together to preserve public media for the future.”
To encourage participation – players of cell-phone and online games like Words With Friends will know the compulsion — players of FIX IT rack up points on a game leader board; scoring is based on identifying and correcting errors displayed through a game portal. Each error a player fixes earns a point, and the corrections become available in the Archive, which is public-media’s largest.
The game borrows an approach from a variety of precursors, including ”fansubbing,” in which fans translate dialog from a foreign-language film or television program and create a subtitled version. The practice, which is of disputed legality, blossomed clunkily in the VHS-tape era, thrived particularly in anime-film fandom, and has been greatly technologically facilitated by the advent of online technologies.
Less legally doubtful are subtitles and captions produced to make audiovisual materials accessible to people with disabilities. The 1996 Chafee amendment to the Americans with Disabilities Act of 1990 — now Sec. 121 of the Copyright Act — affirmed as much.
And the amendment, along with subsequent court rulings, has affirmed that educational organizations can create captioned versions of audiovisual materials for educational use without risk of copyright infringement.
Of course, translations and captioning done with copyright holders’ consent risks no legal vulnerability, at all. That protects entities like Amara, an online community where volunteers can use an Internet platform to transcribe, translate, and subtitle videos for nonprofit and other organizations. Amara provides free, open-source tools, and a platform, for collaborative contributions.
A project like FIX IT falls in the same category, as the transcriptions it entails are from copyright-owning public stations affiliated with the Archive.
— Peter Monaghan
Previous Post: The Curious Case of the Disappearing Video Archive