Let’s face it, the shape of the language industry is changing rapidly. There is a growing need for translation services as companies expand to reach a global market creating a flood of material to be translated from product development to deployment. Digitalized content is being created continuously in places like social media and e-commerce sites worldwide, and the audiences for these are far reaching as well. The demands placed on the industry have challenged the role played by linguists when delivering language services.
Increasingly more companies are relying on machine translation to speed up the delivery of language services, however, to compensate for the quality that MT lacks, post-editing services are more important than ever. The first challenge in the post-editing environment is to work with a variety of software tools in use for machine translation.
Post-Editing Vs. Translation
Post-editing requires a different approach than straightforward translation. If you have ever had the task of editing a document written by someone who is not a native speaker of a language you understand the challenge. Much like if I were to carry on a conversation in a language that I am not quite fluent in, sometimes the word order, parts of speech, etc. just aren’t quite right and don’t flow smoothly. It can be difficult not fall into the trap of keeping poor sentence structure and awkward lexicon while keeping the meaning of translation intact.
The challenges with post-editing MT output can vary widely on the individual type of machine translation used. Therefore, the skills and techniques are required for post-editing change depending on the type of content. Rule-Based Machine Translation (RBMT) and Statistical-Based Machine Translation (SBMT) often create different challenges for a post-editor to look for. For example, an RBMT output may have the correct words translated, but the word order could need drastic improvement. Alternatively, with a statistically based machine translation, the translated material may be too far from the meaning of the source content.
TAUS, Machine Translation Post-Editing Guidelines, differentiates two types of post-editing, full and light post-editing and the importance of determining quality expectations for each. When a high-quality translation is required, full post-editing is needed however when the quality is not the top priority, a light post-editing may suffice. While both should look at the grammatical translation, a light post-edit does not always require work on the overall flow, style, and readability of the text.
Machine translation and its use across multiple industries and platforms is on the rise. Increasingly more developments are being made which are drastically improving the output of MT and companies are investing in these digital solutions. New neural machine translation (NMT), which combines the power of machine translation and machine learning, is changing the landscape of translations as we know it. This again presents unique opportunities for linguists as not only the human-touch point of post-editing but the translation of the data being utilized by these neural machine translations.