Bilingual e-books via neural machine translation and their reception

This chapter discusses the application of neural machine translation (NMT) to works of fiction that have been published as e-books in a bilingual setting. These bilingual e-books are translated using NMT for readers with an intermediate or high level of a foreign language who want to read novels originally written in English. NMT is used to provide a translation in their mother tongue that the readers can refer to on demand, e.g., to help them understand difficult passages in the original. The main hypothesis of this research is that the quality provided by tailored NMT systems is sufficient to meet this goal and that using this methodology will enhance the readers’ experience. We expect the users of these bilingual e-books to read most of the book in the original language and to use the translation only sporadically when encountering difficult parts. This could help the reading experience become more intense, since the reader does not need to consult external sources such as dictionaries or use external public NMT engines.In this chapter, we first discuss our motivation and present the goals of the project. Then the previous work in this field is presented. After that, we explain the process of customizing an NMT engine and creating the bilingual e-books. We follow with the design of the survey and the results of a reading experiment using an English short story set up for Catalan native speakers with a between-subjects design. Finally, we reflect on the results and propose avenues for future work. We are mainly interested in describing a useful methodology for creating bilingual books with NMT and assessing their usability by involving L2 readers.