Having known just a little about the development and integration of closed-captioning technology, we really appreciated this fascinating deep dive from Radio Lab into its history and struggle for equal access that followed, with accommodation, advances in hardware and software, representation and mandates all intertwined and informing one another, concluding with a reflection on how the process is being automated with artificial intelligence and how in training the machine, we ourselves are transformed through the collaboration. Of course the story didn’t end with triumph of accessibility through the above first demonstration, as the advances for the hearing impaired community were not widely accessible: most programming was not captioned and for those that were an expensive decoder was required as a television peripheral. The situation gradually improved and after the passage of the Americans with Disabilities Act, TV sets were required to include closed captioning technology and all broadcasts were mandated to include subtitles. 
A workforce of thirty thousand transcriptionists were at work to capture all stations’ content and in order to reach all of the growing market with the rise of cable programming, institutions providing the service turn to emerging voice recognition systems. These early versions were too bug-prone to be useful, especially for realtime applications and failed to keep pace with live dialogue, seizing up at the slightest accent. Researchers, however, discovered that they were more responsive and accurate with the voices of the trial participants, and soon one devised helping the computer by reading back the words in a steady, well-enunciated manner that it could manage. A team of voice writers across the States repeated scripted shows and news reports as they were aired and achieved a pretty good level of fidelity by 2003. Even with only their master’s voice, the programme still had its shortcomings and the voice writers developed a code of substitute words to clear up homophones and short prepositions, for example: echoing, “She has tootoo daughters inly college comma tootaloo period” would yield the yield the desired text, “She has two daughters in college, too.” Two decades on, the software has advanced to the point where it can transcribe instantly without the help of an interpreter and is improving with AI refinements.
As Boing Boing informs, on this day in 1974, Donald Sherman, who had Mรถbius Syndrome, a rare congenital disease that results in facial paralysis, and had the inability to speak, was able to order a pizza by placing a call from the Michigan State University’s Artificial Language Laboratory. The revolutionary text-to-voice synthesiser (see also) was designed by university researchers and the successful exchange was captured for posterity by local media, though it didn’t go off without a hitch as the synthetic voice was unexpected by the operators—with major delivery chain Domino’s hanging up on the caller—until a sympathetic employee at a small pizzeria took the order. Celebrated annually on campus, Domino’s has been furnishing free pizzas for the commemoration, ostensibly out of the bad publicity for hanging up on Sherman all those years ago.
