Sometimes a story in the trade press sparks a sentimental re-examination of current events. My newsfeed for stories tagged “speech processing” delivered this story from “Broadcast Engineering” Magazine. The headline reads: “UK teletext specialist SysMedia launches new subtitling software.” For those readers who don’t know “teletext” or “line 21 of the Vertical Blanking Interval,” let me just say that this 35 year-old technology was probably the first delivery scheme for “on-demand” delivery of news and information displayed as text on a TV screen. It pre-dates home computers and is a precursor to RSS feeds and other Internet-based news feeds.

Here’s the best part. SysMedia, which has been in the business of providing closed captions and sub-titles for 30 years now is incorporating speech recognition into its process, but… as it explains here, it “does not mean recognising the voice of the TV presenter – the technology is not yet good enough to deliver sufficient accuracy.” Instead, SysMedia uses speech recognition as a form of triage. It performs automated transcriptions to trained “respeakers” who talk into trained speech rec systems for more accurate transcription.

There are other enhancements, of course. The system is optimized to control the speed that subtitles scroll across the TV screen. Speech recognition can also be used for offline preparation of subtitles or captions. This also supports the addition of time codes.

Categories: Articles

Tags: Automated Speech Recognition, information retrieval, Interactive TV