In “2001: A Space Odyssey,” HAL, the computer, can speak and play chess. By 2001, computers could play chess at grandmaster level, but even today, Siri isn’t quite as fluent as HAL. Human language is proving more mysterious than chess to the silicon monster, which is reassuring to those of us whose chess game is a little short of grandmaster level. Another type of computer language use is automatic translation—one of the first applications of computer power with a long history.

Automatic translation, also known as machine translation (MT), is about using a computer to translate from one language (the source) to a second language (the target).

The latest MT technology development stage, neural machine translation, can produce text that is usable in many contexts but still short of impeccable quality—if you have used Google Translate, you might have noticed an improved quality in recent years.

So if you want to translate your website automatically, you can put your source text through an automatic translation app and get an initial output, but it most probably won’t be perfect. You might be able to use Google Translate for a text and obtain something comprehensible, but it wouldn’t be of sufficient quality to publish or put on your website—plus there are many other accompanying tasks involved. These are in the realm of automated translation.

Is automatic and automated translation the same?

Imagine one of the first chess computers: The famous Turk, an automaton created in the late 18th century. Modern chess computers will just spit out the next move unceremoniously on the screen. The Turk was dressed in Ottoman robes and turban, smoked a long pipe, and moved the pieces on a real board for you as its mysterious mechanical workings calculated the variations. It would also knock the pieces off the board if you tried to cheat. You got a real chess experience with the Turk.

Likewise, there is more to automated translation than automatic translation, which is just the equivalent of giving you the next move. There’s a difference between a computer saying “0-0” or “d4” and seeing the actual pieces moved, especially for the end-user. Your website will have text, links, pictures, a special layout, maybe even videos, and there is more to translation automation than just feeding it through an automatic translation app.

Your price list for the USA might be in dollars, but when it goes into French, for example, it needs to be displayed in euros (and take into account an exchange rate, not to mention using metric measurements instead of US ones). You might have a picture of your product being used in Time Square in the US section of the site, with a suitable caption. Both the picture and caption would need to be adapted for the French version of the website.

Automated translation refers to automated parts of the translation process—in particular computer-assisted translation (CAT) and automated translation management.

CAT tools handle the human translation environment, providing matches with sentences and terms that have been translated before, and can also be linked to automatic translation tools. They can include glossaries or term bases, which, in turn, can be connected with machine translation engines. You can tell your MT engine in advance that you want to use the term “batch” and not “lot”, for example, or provide it with an MT glossary of your in-house technical terms.

A translation management system can perform repetitive translation-related tasks through integrated triggers that let the system know what can be automated. For example, commonly used text such as legal disclaimers can be inserted into documents from a content management system (CMS). Automatic translation could be used as part of this automated translation workflow—but the two concepts are distinct.

Automatic translation vs machine translation: What’s the difference?

Automatic and machine translation means the same: The process of transferring content from source to target language without any human input. It’s part of automated translation management.

In terms of the Turk, automatic translation is the bit that works out the best move, the automation is the rest of the apparatus that sets the whole thing in motion giving the end-user a real experience.

Probably at this point, we should admit that although the whole mechanism was an impressive array of levers, pulleys, and magnets, the actual brain was a chess master secreted inside—but those were early days, and things have improved since then.

How does automated translation work?

A translation management system is the motor of automation in translation. Automating your multi-language operations will save you time and increase quality while reducing time to market:

• It lets you automate your workflow through an API, meaning you can deliver faster and reduce error.
• Repetitive manual work is eliminated as you can manage across other languages and locations.
• It enhances team collaboration—staff from all other parts of the world are connected through a single tool in a streamlined collaboration.
• Meandering email threads become a thing of the past.
• A robust translation management tool provides a solid translation memory. This is especially important when working on multiple projects that use common terms.
• You can speed up productivity by storing all past project information that your translators can recall with the touch of a button.
• The TMS can set up new projects automatically as it detects new content, insert previous translations from existing memories, calculate the cost of each translation project, and send out email notifications to language service providers (LSP) or freelance translators who can be chosen on a first-come-first-served basis or selected based on previous experience.

What are the benefits of automatic translation?

It might be fair to say that automatic translation approaches the standard of human translation asymptotically. Far better than it used to be, under its latest form, neural machine translation, it is no longer prone to risible errors. It can be trained: As you generate more output, your MT engine can learn and improve.

For the best result possible, machine translation needs post-editing that is checked by human eyes—arguably this is what the translation profession is becoming (if you remember the chess player hidden in the Turk, we haven’t quite gotten rid of them). At the same time— ironically—solid machine translation is sometimes even more difficult to proofread or edit because the shortcomings aren’t so glaring and become harder to spot.

Should you use automatic translation on your site?

Using automatic translation on your site will greatly depend on the language combinations you need and the type of content you have. Technical and legal texts will benefit more from machine translation, while customer-facing marketing content may not fare so well.

In general, you can follow these 3 key rules when you need to decide on using machine translation:

• Use raw machine translation for low-impact and unambiguous content (e.g., internal documentation, instruction manuals, chat/email support messages, website footers, etc.)
• Apply light or full post-editing to more sensitive content (e.g., product titles, product descriptions, knowledge bases, FAQs, etc.)
• Stick to human translation when branding and culture come into play (e.g., homepages, landing pages, newsletters, press releases, advertising banners, SEO content, etc.)

Localization and Translation Glossary

Stay current on the latest terminology and advancements in the localization industry—so you don't miss a thing.

Browse the glossary

What’s the best translation software?

An understanding of automatic and automated translation puts you in a position to assess your options for the best translation software. A modern, cloud-based translation management system can automatically create a project if new content is detected. The next possible step is pre-translation where translation memories and machine translation insert translations before any human translators even start working. Certain types of content will benefit particularly from this feature.

At this point, translators and vendors can be notified of new work, or the translation software can even suggest the optimal linguist for the project according to the required expertise, say legal, technical, or more creative. The financial costs can be calculated as soon as the job is created, and jobs can be assigned automatically on a first-come-first-served basis to translators or MT post-editors. You may only have to make the first move, when you choose your translation management system, the Turk can do the rest. Try a Queen’s Gambit maybe.