“1996 DARPA/NIST Continous Speech Recognition Broadcast News Hub-4”
The DARPA Hub-4 test that began in 1995 is evaluating the use of large-vocabulary continuous speech recognition (LVCSR) to transcribe audio recordings of broadcast news. Several preliminary Hub-4 evaluation results have been reported [1-5]. Coincidentally, in 1996, the Japanese government announced that it will issue a regulation in several years requiring TV news programs to be closed captioned.
Targeted broadcast mediums in initial inquiry(not in order of importance)
- CSPAN Washington Journal
- PRI Marketplace
- ABC Prime Time
- CNN World View
- NPR Morning Edition
- NPR The World
The delivered project would need to be able to take in SPHERE-formatted waveform files and deliver a text transcription. The Japanese market at the time quickly mandated that all broadcast media had to include closed captioning. Since then we have seen similar movements in the United States market with a continual march towards everything have a text transcription(alt tags on images, digital broadcast of television) under the name of handicap accessibility. The reality that all data parsing tech is VERY similar to the methods that visually impaired persons use to process the world.