How to get your favorite translation

Neural machine translation is attracting attention because it can generate very fluent translation compared with conventional statistical machine translation. However, since neural machine translation learns the probabilistic model expressing the relationship between the source sentence and the target sentence only from the parallel bilingual text, there is a problem that it is difficult for the users to finely control the sentence output by the translation system. In this exhibition, we show how users can customize the output of neural machine translation systems by using tags which represent arbitrary features of the sentence output.

Please click the thumbnail image to open the full-size PDF file.