If you pay attention to news from the translation and localization world, you are almost certainly aware that machine translation has become increasingly sophisticated in the past few years.
Using machine translation (MT) for technical documents might sound appealing. After all, technical documentation translation services can be costly: documents are often long, they require specialized translators, and they may need to be translated for multiple markets.
Note: In this article, we use the phrase “technical translation” as a catch-all. IT, manufacturing, and telecommunications are examples of what we would refer to as “technical” fields. Within each of those fields, you can get more specific; automotive manufacturing is very different from textile manufacturing.
Despite advances in MT technology, however, we do not recommend it for documentation translation services. Technical manuals, fact sheets, MSDS, and other documents should still be handled by a human translator.
Let’s take a look at several reasons why.
Machine translation tools are only as good as the dataset used to train them
To produce a high-quality translation, a machine translation engine needs to be trained. It does so by consuming texts that are available in both languages.
Imagine that you want to train your machine translation engine to work on automotive texts in the Chinese into English pair. You will need a dataset that consists of that type of document, in that language pair. This may work if you already have a database of similar files, and you are confident that the translations are high quality. If not, it can be difficult to find enough documents that meet both criteria.
For technical documents, an all-purpose machine translation engine may be inadequate. For example, an engine trained using minutes from the European Parliament is not suitable for translating drivers’ manuals.
Even if you have a dataset on hand, or you can find one to purchase, there are other compelling reasons not to use machine translation.
A successful technical translation addresses a specific audience
As part of our documentation translation services, we start by asking our clients specific questions about their target audience. Who will be reading the translation? Where do they work? What experience do they have? Are they a member of the general public or an industry professional?
Using this information, we then prepare a brief for the translator and editor that will inform the decisions they make, especially with regard to terminology.
Two of the major factors we consider are:
- Expertise: A one-page brochure, created for sales representatives to hand out at a trade show, will use general language to describe a product. The manual for the same product, aimed at the person in charge of maintenance, uses more technical terms.
- Target market: When a language is spoken in multiple regions or countries, each may have its own region-specific vocabulary. As an example, Spanish speakers in Mexico, Spain, and Argentina all use different words to refer to the trunk of a car. This applies to technical jargon as well as words used in everyday speech.
A translator will be able to identify nuances and do the necessary research that a machine translation engine is not capable of at this time.
Technical translations need to be adapted for the target market
Not only does a translator select the correct vocabulary for the target market, but they make other changes as well.
In some cases, they explain terms that might be unfamiliar to the target audience. In other cases, they may leave out words or phrases that are redundant when translated.
For example, news articles in the United States often use the phrase “the White House” to refer to the administration that is in charge, or its spokespeople. Most American readers understand this shorthand. But a literal translation might not be clear to a Brazilian reader, who only knows that phrase as the name of the building itself. A skillful translator adjusts the text to read “the president’s staff” or finds another equivalent that works within the context.
A translator familiar with the target culture also flags words, phrases, or imagery that might cause offense. What is acceptable in the source culture may be problematic in the target culture.
Take the words “wearing a green hat.” In English, they do not carry any special meaning. But in Chinese, that phrase can be used to indicate that a woman has been cheating on her partner. If the term were translated as-is, the result could be quite offensive to a Chinese audience.
Likewise, if there are culturally specific elements that need to be modified to help the target audience understand, a translator can suggest those changes.
Human translators make clarity a priority
A successful technical translation is ultimately an example of good technical writing—and good technical writing is all about clarity. In a well-written technical document, each sentence can be interpreted only one way.
While working on a technical translation, a human translator may find ambiguities in the source text. They then ask questions so that the target text does not reproduce the same ambiguity.
Once the first draft is complete, an editor revises it to make sure the translator has not introduced any ambiguities into the text.
Let’s look at a sample sentence in English:
I can’t cut that tree down with that axe. It is too small.
The word “it” could be referring to the tree or to the axe—it is unclear. A translator looks at the context of the sentence, and if it is still unclear, they can consult with the writer or client contact.
A machine translation tool does not have that capacity, so it simply chooses one option.
We translated this sentence into Spanish using an MT engine.
No puedo cortar ese árbol con ese hacha. Es demasiado pequeño.
The adjective pequeño (small) is gendered in Spanish, so here it clearly refers to the tree (el arból, masculine) and not the axe (la hacha, feminine).
There is no ambiguity in the Spanish sentence—the tree is too small to be cut down. But depending on the context, that may not be the intended meaning.
Context is key to documentation translation services
Any translator knows that even though a phrase is short, it can be challenging to translate. That is especially true of any phrase that appears without context since it can have more than one possible meaning.
When a human translator finds a phrase or string without context, they can check it against the source document to find where it occurs. They can also review other translated material for possible solutions.
A machine translation engine, on the other hand, makes an educated guess based on its own data.
For example, there are many English words that are used in other languages when talking about IT. When discussing cybersecurity threats in Brazil, Portuguese speakers use the word “worms.” But even when the context of the sentence is clear, machine translation may not pick up on it.
If a person reads the sentence “Worms can modify and delete files” in English, they know that the worm in question is malware and not an earthworm. But we tested the sentence in a popular MT engine, and it translated the word “worm” literally into Portuguese, instead of leaving it in English as an IT expert would do.
Another example is the use of “precautionary statements.” In the United States, for example, manuals and labels that adhere to OSHA and ANSI standards use four types:
- DANGER will result in serious injury or death
- WARNING could result in serious injury or death
- CAUTION could result in moderate to a minor injury
- NOTICE could result in property damage but no physical injury
Each of these words has multiple possible translations in any target language. If translated without context, the word selected might not connote the same level of danger in the target language as it does in the source.
One final example: one of our French translators was once given a machine-translated glossary for a medical device manual. They quickly identified a major error: the MT engine had translated “lead” as plomb, the chemical element/metal.
But in this case, the word referred to the electrical circuit determined by the position of the two electrodes of the electrocardiogram. When given just the single noun, the machine had to guess which of the two homonyms was being used—and it guessed incorrectly.
Selecting the right translation is critical for both usability and future liability.
Good technical writing impacts your company’s image—and more
Dr. Sue Ellen Wright, an expert on technical translation, sums it up:
“In the modern commercial/industrial environment, high-quality documentation implicitly communicates an overall, company-wide commitment to high-quality products, user-friendly operation, and responsive customer support.”
In other words, quality documentation increases a buyer’s trust. That is especially true for costly and complicated products.
Not only does good technical documentation improve your brand image, but it is also vital for legal reasons.
In Canada, consumer products must have bilingual (English and French) labels, and products that are sold in Quebec must have complete French-language documentation. Similarly, in 1998, the European Union passed a resolution stating that product documentation must be translated into the language of any country where it is sold. Even in countries such as the United States, where translated documentation is not mandated, manufacturers need to consider whether their documentation adequately protects against lawsuits.
Additionally, distributors and other partners may have their own requirements when it comes to technical documentation.
Well-translated documentation opens many doors when it comes to exporting your product. It also enhances user experience and makes your brand seem trustworthy.
Art One’s documentation translation services start with a team of carefully chosen translators and editors. Our project managers field questions from the translation team and then solicit input from the client to ensure that we get the terminology exactly right.
Contact us to learn more about our technical documentation translation services.
Comments are closed.