U.S. Translation Company offers both gisting and PEMT.
Gisting is raw machine translation output used to get a general idea of what a text is about. Full PEMT, or post-edited machine translation, involves editing machine translation output so that the end result is similar to a human-translated text. However, both of these machine translated solutions should be used only in certain contexts. The main considerations are content type, language pair, and intended audience of the target text.
U.S. Translation Company clients typically use raw, unedited machine output for internal use only. From gisting, they are able to get the, well, gist of a text. While gisting does not produce a document of publishable quality, it does communicate the primary message. This works perfectly for internal communications, databases, and the like. Gisting can also allow a U.S. Translation Company client to assess whether they want to have the document fully translated. Clients typically do this with requests for proposal (RFPs), tender requests, and similar content. By using raw machine translation in such cases, clients can save up to 70% of what full human translation and desktop publishing would have cost them. In addition, raw machine translation is fast —so fast, in fact, that clients often experience same-day turnaround.
As the term suggests, PEMT involves a post-machine process. A human editor checks for such elements as consistent style, proper spelling and grammar, and cultural adjustment; as well as basic formatting and other quick quality assurance protocols. Essentially, the PEMT text goes through a lighter version of our regular human translation and quality assurance process. U.S. Translation Company clients typically utilize our PEMT for manuals, software, and other such content in which cultural and linguistic nuance are less relevant. Our clients save up to 30% by utilizing PEMT over standard human translation and full desktop publishing, while still receiving a quality level that approaches that of our standard translation process. For a more in-depth explanation of the process, see Austin Becker’s blog post, Not Your Grandfather’s Translation.
PEMT and Content Type
Machine translation works best with technical, informative documents, and not so well with creative, descriptive texts. Ideal texts for machine translation include legal, scientific and medical texts. Documents with specific and consistent terminology, standard grammar, and recognizable patterns. Conversely, creative and descriptive texts confuse the translating machine with their ambiguity, unconventional grammar, idioms, double meanings and implied meaning. As of yet, machines lack the nuance and sophistication to successfully process such linguistic complications.
PEMT and Language pair
Machine translation has more success with some language pairs than with others. The reason for this discrepancy is the asymmetry of research into some languages over others. Computational linguists have invested considerable effort into certain languages, less in others, and none at all in still others. In addition to the research factor, a larger body of linguistic data exists for some languages than others. Machines, after all, must be trained, and data is the input with which they learn.