Should Scholars Trust Machine Translation of their Articles?
New Books in Science, Technology, and Society
00:00
Neural Machine Translation - The Most Recent Paradime
With the advent of internet and all these parallel athat translators had actually produced some mono lingual data, but mainly a lot of by lingual data created through translation memories and catols. Then there was available a lot of digital data that was bylingual and that was parallel. So applying some statistical measures to this, researchers said, oh, we could apply different, different m formulas for each language. But it resembled a set of rules the new programme for each language and its equivalente. The results were quite poor, and it required a lot of work to get these rules upon running. And what was better was that it was much easier to get an engine up and running quite
Play episode from 05:42
Transcript
