Neural machine translation (NMT) causes quite a stir among language service providers, and it will continue to change the translation and localization industry in the coming years. With the use of large data sets of translation to train artificial intelligence and machine learning models, the quality of translation provided by NMT has greatly improved. More importantly, when NMT is combined with human editors, the end result of translation is top-notch, both technically and culturally. As a result, NMT plus human post-editing is in high demand.

Different Types of Machine Translation

Machine translation automatically transfers content between two languages through a variety of models, including:

  • Rule-based machine translation (RBMT), which produces translated text by creating grammar rule sets for both the source and target languages.
  • Statistical machine translation (SMT), which uses statistical models from the source and target languages to generate translations.
  • Hybrid machine translation (HMT), which builds on the approaches from RBMT and SMT to create translations.
  • Neural machine translation (NMT), which relies on a large artificial neural network to calculate the probability of a sequence of words to generate translations that are contextually accurate.

By using a machine translation service, you can expect a quick turnaround, a high production rate, and efficiency throughout your translation project.

Advantages of Combing NMT with Human Post-Editing

While NMT has clear advantages, including efficiently translating languages that are grammatically intricate, adding human post-editing brings a critical component to the translation process. During this step, a professionally skilled human translator carefully reviews the machine translated content to ensure that the text is well suited and accurate for specific cultures, regions, languages, or specialties. As a result, your translated content will resonate with your target audience.

There are three categories of human post-editing to choose from, with each one offering an important service to your project:

  • Light post-editing, this primarily involves the translator focusing on making sure that the content is readable in the target language.
  • Full post-editing, which focuses on ensuring that that the content is precise and understandable, and that it flows well.
  • Human-level post-editing, which produces content that reads like it was originally written in the target language.

Overall, using machine-aided human translation can guarantee the best translation output for your project.

Steps to Make the NMT + HE Process More Efficient

When getting your content ready for the NMT plus human post-editing process, here are some considerations that will save time and money on the translation project:

  • Focus on Terminology Management

Terminology management is a great way to make sure that your standard terminology is accurately translated. By identifying and storing key terms for your product, customer, or company, you can be guaranteed that any unique jargon is not lost in translation. Additionally, technology management can be used as a means to store words that should not be translated. With the help of computer aided translation (CAT) tools, you can develop standardize terminology databases for use in your translation project.

  • Count on a Translation Management System

By automating the translation process, a translation management system (TMS) reduces manual steps and eradicates repetitive tasks. Using a good TMS is vital to managing large volumes of translation work, so it is especially helpful with a complex translation project. Since translation projects often require teams of people in different locations to work together, a TMS can help ensure that the project is managed efficiently, with everyone staying aware of what others are doing on the project.

  • Use Translation Memory

Translation memory, often a part of the TMS, maintains records of formerly translated fragments and phrases. By ensuring the translation memory is updated for the target languages, you are taking another step toward making certain that your translation output is completely accurate. Having a translation memory is particularly helpful when translating technical documents, since they typically include specialized terminology and repetitive phrases.

Huge companies, such as Google, have seen firsthand how advantageous NMT can be, especially considering its ability to be trained in both the source and target languages. On the other hand, in a recent poll of professional linguists, almost half of the respondents stated there is a huge increase in the number of clients who are requesting human post-editing. Considering the value of both NMT and human editing, it is clear that a combination of both approaches will provide the best possible translation output.