Process Automation for Localisation of Dialogue in Entertainment Media (PALODIEM)

This project concerns the exploitation of latest academic research in the field of natural
language processing to develop innovative workflow automation solutions for the
entertainment industry and other sectors in which localisation of audio/visual materials is
important.
The project will focus on localisation of soundtracks, captions and subtitles and will use cloud
computing to improve process efficiency, enabling more effective localisation of crossplatform
digital media for feature film, TV, video games and other non-entertainment content
such as news, education and corporate communications.
The applicant is ZOO Digital, a company that develops and markets cloud-based workflow
software for creative organizations. ZOO already provides subtitling services and, in the
course of creating subtitles for entertainment industry clients, is building a large proprietary
database of dialogue data with parallel translations into over 50 languages. For this project,
this database will be used to ‘train’ machine translation software to enable a capability to
suggest translations of new dialogue, resulting in faster turnaround and lower costs.
Quality estimation methods will be used to develop optimum strategies of capitalising on the
similarities between related languages (e.g. Romance languages - French, Spanish, Italian,
etc.), where the approaches are expected to give superior results. Workflow management
methods will be developed for increasing productivity of multi-lingual dubbing, from casting
through to editing and mixing. Automation will be used extensively to enable high levels of
reliability with low human intervention.