Neural Net Machine Translation May Increase Acceptance by Human Translators
January 2, 2018
Apparently, not all professional translators are fond of machine translation technology, with many feeling that it just gets in their way. A post from Trusted Translations’ blog examines, “Rage Against the Machine Translation: What’s All the Fuzz About?” Writer Cesarm thinks the big developers of MT tech, like Google and Amazon, have a blind spot—the emotional impact on all the humans involved in the process. From clients to linguists to end users, each has a stake in the results. Especially the linguists, who, after all, could theoretically lose their jobs altogether to the technology. We’re told, however, that (unspecified) studies indicate translators are more comfortable with software that incorporates neural networking/ deep learning technology. I seem such tools produce a better linguistic flow, even if some accuracy is sacrificed. Cesarm writes:
That’s why I mention emotional investment in machine translation as a key element to reinventing the concept for users. Understanding the latest changes that have been implemented in the process can help MT-using linguists get over their fears. It seems the classic, more standardized way of MT, (based solely on statistical comparison rather than artificial intelligence) is much better perceived by heavy users, considering the latter to be more efficient and easier to ‘fix’ whenever a Post-Editing task is being conducted, while Post Editing pre-translated text, with more classical technology has proven to be much more problematic, erratic, and what has probably nurtured the anger against MT in the first place, giving it a bad name. Most users (if not all of them) will take on pre-translated material processed with statistical MT rather that rule based MT any day. It seems Neural MT could be the best tool to bridge the way to an increased degree of acceptance by heavy users.
Perhaps. I suppose we will see whether linguists’ prejudice against MT technology ultimately hinders the process.
Cynthia Murrell, January 2, 2018
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
