Machine Translation technology has advanced significantly in the past decade, but not to the degree to not require the involvement of the human factor. And here comes our word of the day, the post-editor, to edit and correct the machine generated output.
The job title derives from the Latin verb edit ‘to produce, put forth’ and the word-forming post- ‘after, behind.’ The sense of editor as ‘a person who prepares written matter’ developed in the 1710s, and the job occupation of one who alters or rearranges computer generated data appeared first in the 1950s.
An MT post-editor is a professionally trained translator or an expert linguist who corrects and improves the raw Machine Translation output, opposed to an MT pre-editor who corrects mistakes in the source text and, often also adapts the source text to help the MT engine produce higher accuracy.
History of Machine Translation started back in 1947 with a letter from Warren Weaver about the possibilities of using a newly invented computer to translate natural languages. Followed by his Translation memorandum, a set of proposals for computer-based Machine Translation, presented two years later.
Weaver’s memorandum inspired the Austrian comparative philologist Erwin Reifler, a Chinese specialist at the University of Washington (Seattle) who has circulated a couple of papers proposing the use of human editors before and after the translation process, to put forward the first formulated conceptions of pre- and post-editing in the early 1950s.
“Reiffler suggests the use of a pre-editor who need not know the (target language) at all but who, reading his own language (our source language), eliminates from the text all such alternatives… In the same way a post-editor might be used to turn the translation into acceptable T.L. “(Andrew Donald Booth and Kathleen Booth, Automatic digital calculators, 1953.)
At the first conference devoted to the topic of Machine Translation, convened at the Massachusetts Institute of Technology in June 1952, Yehoshua Bar-Hillel – an Israeli mathematician, and linguist, gave a paper entitled MT with post-editor, stating that Reifler’s proposals were not feasible and proposing that the functions of pre-editing could be virtually eliminated and then only a post-editor would be necessary. Bar-Hillel believed that while the post-editor must know the target language, it was not necessary he to understand the source language.
First in-depth studies on post-editing appeared in the 1980s, and today, nearly 40 years later, when more than 50% of all Language Service Providers and individual language professionals report that they are using Machine Translation in one form or another, post-editing is required in almost every instance where MT technology is used.
Translation buyers can choose between two levels of post-edited machine translation – light (taking the raw MT output and performing as few modifications as possible, usually transferring the correct meaning, while ignoring any stylistic issues), and full (meeting the quality criteria for human translation).