The deep turnaround in machine translation

Speaker:
Ondřej Bojar
Abstract:
In the talk, I will summarize the major shift of paradigms that happened in machine translation (MT) in 2015-2017, leading to neural MT (NMT). Morphologically rich languages no longer seem to cause a distinct challenge and the errors made by current best MT systems are getting closer to errors made by humans. I will present the current best approaches, their shortcomings and research topics relevant for the near future. On a general note, the relation between linguistics and engineering in MT has been disrupted by NMT and, to some extent, 'we will again have show the importance of linguistics' for MT. Conversely, I am trying to find out what NMT can offer to linguistics. I would like to argue that NMT provides a great object of study, a 'dissectable language learner'.
Length:
01:40:17
Date:
30/10/2017
views:

Images:
Attachments: (video, slides, etc.)
138 MB
167 downloads
1.1 GB
176 downloads
593 MB
182 downloads
309 MB
170 downloads
216 MB
175 downloads