Speech recognition-based subtitling, known as respeaking and used mainly for the production of live subtitles, has enabled broadcasters to increase considerably the amount of subtitles provided for their deaf and hard of hearing viewers. Yet, now that quantity is in some cases no longer an issue, it is time to focus on the quality of these respoken subtitles. Attention may then be turned to the viewers: What do they think about respoken subtitles? How much information do they obtain from them? How do they actually view/read them? The aim of this paper is to answer some of these questions by presenting the results of two studies on the quality of respoken subtitles.
In the first study, included within the framework of the EU-funded project DTV4ALL, a number of experiments were carried out with three groups of participants: deaf, hard of hearing and hearing viewers. After showing the viewers several news clips subtitled by respeaking, a questionnaire was handed to them to find out how much (verbal and visual) information that they were able to obtain from the subtitled news as well as their general views on respoken subtitles. Finally, an eye-tracker was also used with participants from all three groups to investigate how these subtitles were read.
In the second study, the reception of respoken subtitles was tested in a different context. Through collaboration with the subtitling company Stagetext and the National Gallery in London, lunchtime talks at the latter venue were made accessible, for the first time ever, using respeaking. Deaf and hard of hearing viewers who attended the talks were able to read what the presenter was saying as it was displayed in real time on their laptops. Several experiments were carried out to explore issues that had been highlighted as important by the participants in the first study, such as the lack of synchrony between the images and the subtitles. To tackle this issue, a test was run to delay the video signal so that there was synchrony between the images and the respoken subtitles.
The results obtained in these two preliminary studies suggest that, for all the progress that has been made regarding respeaking, there is still room for improvement. It is argued that some of the main aspects to be addressed are the display mode of the subtitles, which, at present, requires time and effort that may be affecting viewers’ comprehension, and the lack of synchrony between subtitles and images, which highlights the need to adopt technological measures to increase the quality of respoken subtitles.

Pablo ROMERO FRESCO completed his PhD on the naturalness of the Spanish dubbing language at Heriot-Watt University (UK), where he taught Translation, Interpreting and Subtitling. He is now a senior lecturer in Audiovisual Translation at Roehampton University (UK), where he teaches Dubbing, Subtitling and Respeaking (subtitling through speech recognition). He has also taught Respeaking at the MA on Audiovisual Translation at the Università di Bologna (Forlì, Italy), at the University of Coimbra (Portugal) and currently at the Universitat Autònoma de Barcelona (Spain), both on-campus and online. He is currently collaborating with the subtitling company Stagetext to provide, for the first time ever, respeaking-based accessibility in venues such as the National Gallery. He is a member of the research group Transmedia Catalonia, for which he co-ordinates, along with Verónica Arnáiz, the EU-funded research project D’Artagnan (DTV4ALL), focused on the standardisation of subtitling for the deaf and hard of hearing in digital TV. He has published several articles on Dubbing, Subtitling and Respeaking and is now writing the book Translation Through Speech Recognition: Respeaking (St Jerome).