Content area
Full Text
Articles
Together with technological advancements, new forms of linguistic mediation are being developed in the area of media accessibility and audiovisual translation (AVT). One such form is respeaking, a method of producing real-time subtitles to live television programs using speech recognition (SR) software (Marsh, 2006; Romero-Fresco, 2011). Respeaking was first used in 2001 by two public service broadcasters: the BBC in the United Kingdom and VRT in Belgium (Lambourne, 2006). Since then, this form of AVT has grown tremendously, and it is now a major method used to produce live subtitling on TV.
Respeaking plays an important role in delivering the original spoken text in the form of written subtitles to people who are deaf and hard of hearing as well as language learners and others who use subtitling to support their TV viewing (Eugeni, 2008a; Romero-Fresco, 2011). In daily life, many of us probably have encountered subtitles to live programs and wondered how it is possible for the spoken dialogue to be converted into subtitles within a matter of seconds. The respeaking process is a complex activity (see Figure 1), in both technical and cognitive terms (Boulianne et al., 2009; Luyckx, Delbeke, Van Waes, Leijten, & Remael, 2010; Romero-Fresco, 2011, 2012). A respeaker needs to listen to what the original speaker is saying in a TV program and to respeak it, that is, repeat or rephrase the text, adding the necessary punctuation marks and important information for viewers who are deaf or hard of hearing, related to speaker identification and sounds. The words uttered by the respeaker are then turned into text using SR software. This text is later displayed on viewers' screens as subtitles with a delay of several seconds (Ofcom, 2015). In some countries, respeakers are also required to correct the output of the SR program if they spot any errors. In other countries, the error correction is done by another person, known as editor or moderator. The error correction process improves the quality of respoken subtitles, but it may increase a delay between the time when the original speaker said something in a TV program and when the corresponding subtitles appeared. Since delay is one of the most frequently voiced complaints by deaf and hard of hearing viewers (Mikul, 2014), every...