Fostering access for all through respeaking at live events

Abstract

Live subtitling using speech recognition, known as respeaking, is widely used to make live television programmes accessible. Although a growing area within audiovisual translation internationally, in the UK its industry use has been limited to television, in part due to the many misconceptions surrounding its production. This study explores how respeaking can be introduced to complement current access provision at unscripted or partially scripted events. Through close collaboration with users and providers, respeaking is shown to be a viable way of providing access for deaf, deafened and hard of hearing audience members in this new sector: access that a wider audience is also likely to benefit from. The paper begins with a brief discussion of the audiovisual landscape, focusing on quality in respeaking and current provision in the sector. Next, a bespoke training programme is presented and user and provider expectations for the service are outlined. Finally, the quality of respeaking at a series of research events is discussed. The results indicate that the quality attained for the most part meets, and frequently exceeds, the benchmark of 98% accuracy set for live television subtitling. Latency is similar to that seen on television, yet remains an area for further consideration.