Abstract
This chapter reports on a project to identify and resolve genre-specific issues in producing subtitles of parliamentary debates for d/Deaf, deafened, and hard-of-hearing viewers/audience members/users. The aims of the project were (i) to use linguistic methods of analysis to generate practical insights for potentially improving both the speed and quality of live subtitling, or respeaking as it is known, and (ii) to test these by respeaking a sample of parliamentary data. Speed is defined as the quickness with which the respeaker can produce subtitles. Quality is defined in terms of latency (i.e., delay between the production of a subtitle and its appearance onscreen) and accuracy. In addition, the chapter explores the possibilities of providing a viewer reliant on subtitles with an experience as close to that of the hearing viewer as possible. This means conveying not just words spoken but associated paralinguistic features too (e.g., pitch, prosody, triggers of implicatures, and politeness features). On the basis of our test respeaking, it is suggested (i) that it would be valuable to develop an extended bank of macros for variations in sound labels, and (ii) that an increased awareness of the possible implications of particular linguistic features is likely to impact positively on subtitling practice.