Last year, Project Manager Jesus Moreno programmed and developed the Stream It video service. Recently, he’s built in IBM Watson closed captioning services, which has proven to be successful.
“[Stream It] is responsive, so if I access it on my desktop or on my mobile device, I don’t have to have multiple formats,” Moreno said. “It’s just one and done. Everything flows.”
For the closed captioning, Stream It uses IBM Watson, which Moreno said is constantly learning accents and phrases and has a very high accuracy percentage.
“Depending on the quality of the audio, we’re getting anywhere from 80 to, I think, we got 96% accuracy the other day,” Moreno said. “It’s pretty robust.”
Producer Michael Coleman also said they’ve met with IBM, who is making constant improvements to the technology.
“That’s what kind of nice about a cognitive engine,” he said. “It learns, but so do the people who are actually engineering it. So for example, right now Watson can’t differentiate between two different voices if both of us were speaking on a video. It’ll translate both, but it doesn’t know you’re Speaker 1 and I’m Speaker 2. That’s something we’re working on.”
Essentially, Watson will grow as Stream It continues to use it, but it will also improve on IBM’s side.
Another advantage of IBM Watson is the captions are timestamped, which makes it easier to go back and make corrections.
“That’s the hard part. So now, all you have to do is go in a fix a bit of the text,” Moreno said. “We’ve found that when we’re dealing with clear audio, the accuracy is very high.”
Coleman said the next phase is to do language translations.
“We’re going to start with five [languages] to see how it goes and then we’re going to move it further than that,” Moreno said.