Neural machine translation (NMT) is a fantastic way to cut translation costs and stretch a budget. When used properly with a technical source text that’s been properly formatted, i.e. edited for conciseness with a standardized vocabulary, NMT can save you somewhere between 30 to 50% of your translation budget. However, different projects require different levels of post-editing which will be reflected in the cost. Today, we are going to look at the difference between light and heavy NMT post-editing to help you decide which is best for your translation project. Raw MT is never ready for distribution. While NMT is brilliant technology that’s revolutionizing translation, the sentence-by-sentence approach lacks paragraph- and document-level context. This is especially problematic with longer, wordier, and less straightforward texts. Furthermore, NMT ignores glossary terminology that is essential for widely distributed or official documents. For this reason, we suggest light or heavy post-editing. Light NMT post-editing ensures a thorough proofread of a MT target text. At this level, language and NMT specialists will check for both lexical errors, for example spelling mistakes and faults in word order, and syntax errors, such as clarity issues and dangling participles. Light post-editing guarantees that a text will be perfectly readable, but it will still sound clunky and machine generated. Heavy NMT post-editing, on the other hand, goes past light post-editing by making a machine translated text sound as if it were translated by a human. At this level, specialists correct less obvious errors, such as ambiguity, and edit the text for appropriate style and flow. This is the most cost effective option for someone looking for a human quality translation of a technical document. Do you still have questions on post-editing options? Visit our NMT post-editing page to learn more!
More posts by Mitchell Thomas