Speech-to-text interpreting – The Road Ahead
Pablo Romero Fresco (Universidade de Vigo)
Speech-to-text interpreting or live subtitling (which allows the production of written access to live events or programmes for people with and without hearing loss) blurs the boundaries between translation, interpreting and accessibility and showcases the tremendous progress that technologies such as speech recognition and machine translation are experimenting in this area. This presentation will focus firstly on intralingual live subtitling. Through four case studies including approximately 17,000 English live captions analysed with the NER model from 2018 to 2023 in the UK, the U.S. and Canada, this presentation will compare the accuracy of automatic live captions to that of those produced by humans captioners. The second part of the presentation will focus on interlingual live captioning or speech-to-text interpreting. It will report on the results of a recent project, conducted with the European Parliament, to compare the quality of different forms of speech-to-text interpreting, from completely human workflows (such as interlingual respeaking) to hybrid approaches (simultaneous interpreting plus automatic speech recognition) to finally completely automatic set-ups (automatic speech recognition plus machine translation). A discussion will follow regarding the current state of the art and the future of (speech-to-text) interpreting and (live) subtitling in light of the recent developments brought forward by artificial intelligence in this area.