A Real-Time MT System for Translating Broadcast Captions
Eric Nyberg, Teruko Mitamura · 1997
This presentation demonstrates a new multi-engine machine translation system, which combines knowledge-based and example-based machine translation strategies for realtime translation of business news captions from English to German. 1. Introduction Broadcast captioning is derived from a textual transcription of a television broadcast, and is typically prouduced in real-time by a human operator using a stenography machine. The ASCII transcription is then encoded and transmitted in the vertical blanking interval (VBI) portion of the video signal. Some broadcasts (such as entertainment shows in syndication), are captioned off-line, but many shows of interest (such as "headline news" programs) are always captioned in real-time, with only a few seconds delay between the audio signal and the transcribed captions. Real-time translation of broadcast captioning poses several challenges for machine translation. The vocabulary and grammar of the source text are uncontrolled. It is not feasible t...