How Organizations Can Cut Localization Costs Up to 40% with Machine Translation Post-Editing

Businesses are under increasing pressure to translate more information quickly and effectively while controlling costs in today's international market. The need for precise and fast translations has increased as businesses seek a wider global audience. Handling these translations without a large increase in expenses calls for creative ideas and calculated approaches to ensure that quality and speed are maintained without going over budget.

Alconost is a professional localization agency that specializes in machine translation post-editing. As such, we couldn't ignore the topic of artificial intelligence (AI) and its implications for businesses that depend on language services—whether they are international offline businesses or web-based apps, games, websites, software products, etc.—as well as for the human professionals who provide these services.

In this post, we'll walk you through the nuances of machine translation and demonstrate how combining human ingenuity with machine efficiency may produce excellent results, cutting down on localization expenses by up to 40% and increasing translation speed by almost 1.5 times! We'll also examine what kinds of text work best with an MTPE process and what works better with a careful, human touch.

How can an MT process save time, reduce costs, and increase the volume of translations you can do? Why is post-editing necessary?

The early machine translation models were far from ideal and frequently failed to capture the nuances of natural language. Thanks to technological advancements, machine translation has become more and more reputable over time.

The industry developed the Post-Editing (PE) practice to address the issue of inaccurate machine-translation output.

This work involves proofreading and editing the machine-translated text by a human translator. Linguists revise the original output to guarantee accuracy, fluency, and cultural appropriateness. In this way, PET contributes to improving machine output quality without sacrificing efficiency.

Connect with an expert

What kinds of content are well-suited for MTPE?

Differentiating between content types and their suitability for machine translation and post-editing is crucial. Generally speaking, traditional human translation is the best option for text with low volume and high nuance. Machine translation solutions can be great for text that is more straightforward and general. Now, let's look at a few instances.

Certain content kinds are still best served by traditional human translation. Because of their intricacy, nuanced undertones, particular context, or requirement to conform to a specific tone or well-defined target audience, these call for linguistic inventiveness. Here are a few instances:

• Advertising

• Marketing content

• Complex legal documents

• Software user interfaces

A combination of human knowledge and machine efficiency can handle other content that is more straightforward and less nuanced. Types of content that fall into this category include:

• User manuals and help information

• Technical marketing materials

• Simple software UI, such as error messages, tooltips, alerts, and notifications

• Financial reports

• Investor information

• Product descriptions

• Support content and FAQs

Finally, some information does not require high quality and is simple to translate. When this is the case, using pure machine translation can cut costs and save time—especially if you are willing to forego human assistance in favor of cost-efficiency. Here are a few situations MT only (raw MT) could make sense:

• User forums and reviews (user-generated content)

• FAQs

• Wikis

• Instant messages and SMS

The following illustration offers a clear and comprehensive guide for determining whether text is appropriate for machine translation.

Since every organization is unique, it's important to remember that there may be particular requirements for every type of material. The members of Alconost's team in charge of the machine translation post-editing stated it this way:

"Remember that there isn't a one-size-fits-all answer in this situation. Every project merits careful consideration. Nevertheless, we've discovered that this service is quite helpful for translating manuals, guidelines, instructions, and internal training materials. However, we would lean more toward human localization when translating user interfaces or marketing materials. It simply adds one more level of dexterity and elegance."

MT Success story: quick, cheap, and high-quality localization in six languages for battery diagnostics

Let's look at a company perfecting the human/machine combination for content localization to demonstrate how diverse businesses may be in content and what they can anticipate from MTPE.

AVILOO Battery Diagnostics is innovating battery data analysis. The business has created a special diagnostic system that offers information about traction batteries for plug-in hybrid and electric cars.

AVILOO needed to translate its various forms of content from German into six languages (Danish, Dutch, French, Italian, Swedish, and Norwegian). Its three goals were to meet the deadlines, maintain high quality of the translations, and work within strict financial constraints.

You may have already guessed that Alconost used an MTPE process for AVILOO's localization initiatives to meet these requirements:

• Initially, Alconost's localization managers employed native-speaking translators to evaluate the output of six distinct translation engines to assess the accuracy of unprocessed machine translations. By following a methodical approach, we were able to identify the best engine for each of the six language combinations.

• Following the selection of the engines, a group of editors who spoke fluent English used their knowledge to improve and hone the output produced by the machine. Their efforts were concentrated on improving readability and removing errors, which produced content that connected with AVILOO's multilingual target audience.

Using MTPE, we were able to complete work around 1.5 times faster than we could have using traditional localization techniques. Also, AVILOO saw significant cost reductions of 30–40% compared to traditional localization efforts thanks to this approach.

With MTPE, AVILOO increased its worldwide reach without sacrificing cost, schedule, or quality. The same is possible for you. Check out Alconost's MTPE offerings or arrange a consultation to discuss your project right now.