Audiovisual production of all kinds is currently growing fast and in parallel with the growing number of people with difficulties in understanding oral texts, paticularly in the West, as a consequence of the fast ageing process and of immigration. Moreover, large public spaces like airports, big shopping centres, sports facilities and conference buildings, among many others, do provide audiovisual live services of programmes and events to all kinds of users in spaces where a high sound level might not be a good solution. In such instances, live subtitling of the visual material is the answer.

The main purpose of QUALISUB is to measure the quality of live subtitles in programmes and events in all four official languages in Spain (Catalan, Basque, Galician and Spanish) as well as in English in the UK, the US and Canada.

This will be followed by a comparative analysis of the quality of live subtitles produced through respeaking and automatic systems in Spain, which, in turn, will be followed by a large-scale NER-based assessment of live subtitling quality in the US and Spain.

Finally, another objective of this project is to develop a technological solution to automatise the NER model. The aim is to issue a number of recommendations that can help improve access to live TV programmes and events for millions of daily viewers worldwide.

To assess the accuracy of speech recognition in Galician, Basque and Catalan, and its potential application to live subtitling.

To produce a white paper with a set of recommendations regarding what is required to provide the first live subtitles in these three languages.

To compare the quality of live subtitles produced through respeaking for TVE, Antena3, Cuatro, Telecinco and La Sexta with that of the automatic live subtitles used by TVE.

To produce a white paper regarding what type of live subtitles (respoken or automatic) may be used depending on the type of programme.

To produce the first large-scale NER-based assessment of live subtitling quality in the US, comparing it to the results obtained in Spain.

To develop a technological solution to automatise the NER model.