02 Sep MTPE: strategies for your business
Have you ever heard of machine translation and post-editing (MTPE)?
Any business that wants to expand internationally needs specialized, professional, state-of-the-art translation services. In the particular case of businesses with an e-commerce site, for example, the accurate translation and localization of all content is critical for building brand trust and loyalty among new customers and for improving the SEO performance of the specific website. In general, we live in a world where the amount of information and content produced is exponentially increasing; what can be done to effectively translate content with reduced time-to-market?
Certainly one viable option for a company is machine translation, meaning translation performed by software, a portal or a website. This solution comes with pros and cons. Machine translation engines can translate up to millions of words per hour, consequently allowing translation costs to be greatly reduced. However, despite the advances made by machine translation engines, the machine’s output continues to have a robotic flavor, making human intervention still necessary to fully grasp the immense and extraordinary amount of facets and nuances of language.
This is why a company must plan a strategy that achieves an optimal balance between speed, convenience and quality, allocating a specific budget to hiring an experienced, native-speaking human translator to reprocess the output of an automated engine, thereby performing what is called machine translation post-editing (MTPE). Machine translation and post-editing services by a human translator are not as fast as the purely automated process and surely incur higher costs in the short term, but they certainly deliver more accurate and professional results, as well as ensuring a positive long-term impact for company image. The purpose of post-editing is to produce good quality translations with the least possible expenditure of time and money, taking advantage of the machine to translate very high volumes of words and intervening only where necessary.
Machine translation and post-editing: a flexible solution
The combination of machine translation and post-editing is a very flexible solution, both in terms of the choice of machine translation engines and the level of post-editing.
- rule-based Machine Translation (RBMT) engines, the first to be developed, are based on a large number of linguistic, grammatical and syntactic rules set by programmers. The output produced by these machines can often be unnatural and unsmooth or have terminology errors, as the software struggles to recognize the specific context of the term;
- statistical Machine Translation (SMT) engines were the next development, in which the best possible translation is proposed based on statistical parameters derived from the analysis of parallel bilingual corpora. The translation is smooth and overall better than that produced by rule-based engines; however, it may have some grammatical or punctuation errors. In general, these engines provide good quality output only if they are trained using extensive, relevant, and above all qualified corpora;
- hybrid engines, which combine rule-based and statistical methodologies to develop different resolution strategies depending on the specific translation problem to be addressed;
- neural engines (NMT, Neural Machine Translation) are the latest and most advanced in the field, exploiting the application of artificial intelligence applied to the field of translation. When faced with the input of large bilingual corpora into the engine, the software autonomously chooses which translation is best and is also able to learn and improve from the feedback of a human translator. Despite their excellent results, these engines still make concordance errors and may have difficulty translating very technical terms or unclear text.
In general, a translator with an excellent understanding of the technology who knows where to intervene in order to be able to improve the machine output, together with a client who is informed and willing to adequately support post-editors, can be the key to choosing and successfully using a strategy that combines machine translation and post-editing. For example, a client who uses statistical engine training (SMT) and provides corpora on dentistry to translate contract texts would get poor or unusable results, and this would require too much effort on the part of the post-editor, resulting in a loss of the benefits in terms of time and money compared to a fully human translation. Another reason why the combination of machine translation and post-editing is very flexible is the fact that post-editing can be carried out more or less thoroughly depending on business needs.
In fact, there are various types of post-editing:
- no post-editing: the automatically translated text is not proofread. If you have almost no budget and a very short timeframe for product delivery, this is the most advantageous solution;
- full post-editing: the text is thoroughly proofread by a human translator who checks, for example, for grammatical, syntactical, terminological and stylistic errors. This is the most expensive and slowest solution, but the final output has very high quality and is potentially indistinguishable from a human translation;
- light post-editing: the text is revised so that the content is understandable, without paying special attention to style or final quality, using the machine output as much as possible. This is an intermediate solution that can have many advantages in terms of both cost and timing.
What factors should you consider when choosing the best combination of machine translation and post-editing?
Determining which machine translation and post-editing strategy is best for your business requires expert help, as well as taking a number of factors into consideration.
The language pair
Machine translation works much better with language pairs that belong to the same language family and share certain grammatical structures. If a company needs to translate from English to German, both Indo-European languages, it will more easily be able to rely on machine translation without post-editing; if, on the other hand, it needs to translate from English to Japanese, machine translation will struggle to handle the linguistic differences and possible ambiguities, making it necessary and safer to have a post-editor to ensure conveying the correct message.
The intended use
If the translation serves merely to understand the source text, then machine translation can be a good strategy. For example, machine translation without post-editing is advantageous for translating low-visibility content such as internally circulated documents, user-generated content or questions and answers within a forum. If, on the other hand, the document is intended for publication and external circulation, or if it contains very specific industry language, the more or less massive intervention of a post-editor becomes essential.
The type of text
Standardized texts and content written in controlled language (characterized by the use of simple, synonym-free, unambiguous sentences and consistent terminology) are particularly suitable for the combination of machine translation and post-editing, as they contain a lot of repetition and are deliberately less stylistically edited. In this case, the choice between light and full post-editing mainly depends on the desired level of accuracy and also on the intended use. Post-editing a cookbook will be less invasive than post-editing a user manual, which in turn will require a different type of attention than post-editing a patent. As the legal implications increase, the importance of accurate post-editing grows in tandem.
At the antipodes, however, we find creative texts such as literature, poetry and advertising content, which thrive on the richness and variety of language. In these cases, knowledge of the target audience’s culture becomes crucial; therefore, it’s preferable to rely on a human translation or, as in the case of the adaptation of a slogan or advertising message, on transcreation, meaning a creative translation that knows how to make the message effective in both the target market and the source market.
Company translation needs are unique. If you are interested in finding out more about the machine translation and post-editing services offered by Creative Words and understanding which technology solution might best meet your needs, please feel free to contact us to find out how we can help you.
Passionate about ballet and singing, she compensates for her years of dieting as a dancer by abusing sweets. Precise and meticulous, she always tries to give her best in what she engages in.