How to customise NMT for subtitling

The use of Neural Machine Translation (NMT) in subtitling is one of the hottest topics of the year, with almost every major provider experimenting with the efficiencies machine translation can provide in subtitling workflows.

I am very proud to have been involved in a case study on this topic conducted on behalf of AppTek. It showcases the improvements that can be achieved in the quality of the NMT output and the productivity gains these can lead to in the work of practitioners, with the implementation of customisation techniques and the use of a novel subtitle segmentation algorithm based on a combination of neural net training and industry-standard rules. More importantly, it tells the story of a successful collaboration between scientists and subtitling experts, and proves that so much more is possible when these stakeholders truly listen to each other.

Read the details here.