Radios, Captioning Algorithm Still Needed - Radio World

Radios, Captioning Algorithm Still Needed

Cost of transcription is among the obstacles.
Author:
Publish date:

NPR would like to establish a regular captioned radio service, but a few things would have to fall into place for that to happen.

The network is trying to cut the cost of live captioning, which can typically cost around $120 per hour, according to NPR VP/CTO Mike Starling, who's also executive director of NPR Labs.

IBM is working on a speech-to-text algorithm that promises to have live real-time translation (rather than 20 minutes or so later). That algorithm is due to be ready in 2009, Starling says.

With such an algorithm — combined with the efforts of, say, an English student cleaning up any remaining garbles at $20/hour — live captioning becomes more cost-effective "and would help the network make the commitment" that it will caption all of its programming next year, he told me.

HD Radio receivers capable of displaying the text would be needed. At least four companies have expressed an interest but none has made a formal commitment so far. NPR plans to host a workshop devoted to captioned radio and all that would entail for receiver makers in advance of CES this year. I've reported that Radiosophy and Delphi are interested; I hear two more receiver manufacturers are as well.

Ibiquity would also need to incorporate the specification for live captioning in its chipsets next year and then work with receiver manufacturers to provide big displays.

Nearly 7 million people in the United States are either deaf or hard of hearing, according to Gallaudet University, and many more have trouble with their hearing.

You know who will benefit from all this work, among others? Us boomers who see retirement on the horizon (or did until the current financial crisis). Quite a few of my friends who were jocks back in the day now have trouble with their hearing.

Related

Cascading Algorithms and HD Radio

While HD Radio is capable of delivering unsurpassed audio, you need to understand some of the issues related to cascading algorithms in order to reap all the aural benefits this medium has to offer.