How Good Is Machine Translation for Global Expansion

A person using machine translation software to test how good is machine translation.

Nowadays, with so many IT developments, learning to make use of the benefits of such progress is critical to discerning the various alternatives available. Not long ago, the question “how good machine translation is” had a vague answer, and many agreed on “not my cup of tea.”

But today, things are substantially different—and it has been the case for some time. This does not mean that translation software will be presented as the answer to prayers of “transcending the language barrier”, but it certainly has many practical uses, especially when combined with human effort and know-how.

The underlying question is no longer only whether this tool is good but if it is suitable for all types of content. If it can solve all kinds of linguistic challenges or whether it is up to the task of international business expansion. Luckily, these questions are the ones that guided us in this blog, so we will squeeze them out in our search for answers.

But there is more, much more. Will you jump on board?

What Is Machine Translation?

Machine translation—or MT—is the process of automatically translating texts from one language to another, through the use of software and without human intervention.

As a discipline, it is a subcategory of computational linguistics, which harnesses developments in computer science, AI, statistics, and more. Pretty complete, right?

And, although it seems a relatively new tool, its first steps date back a long way—you can track them as far back as the 1950s! But, because of data processing power and storage requirements, it is only in the early 2000s that it became kind of a buzzword.

Even at the beginning of the current century, there was still considerable ground to cover, and the first translations were rather inaccurate and with enormous quality flaws. So, the question of how good the translation machine is would keep an implicit negative answer for some extra years.

Is this still the case? Of course not! The landscape has changed and quite a lot! We can attribute this, in large part, to the efforts of Google (which paved the way for other translation engines to emerge as well.) But it is also, specifically, thanks to neural machine translation, which is one of the best-performing models (for MT.) With it, the machine not only translates but also learns while it is in use, constantly improving the quality.

#OptimationalTip: The answer to how good Google Translate is has been turned on its head. However, while the prospect is more encouraging, the machine is not enough to meet QA standards, so you should rely on a thorough MTPE process.

How Good Is Machine Translation: The Pros

From rule-based (RBMT) to statistical (SMT) and from there to neural (NMT): These were the stops on the journey of the MT to become what it is today. The models have evolved so much that Google Translate, as an alternative for business, is a real candidate now. However, this has limited applicability, and its usage should be under careful scrutiny.

But before we talk about it (and look like spoilsport), let us discuss the beauty of it: The benefits and advantages of machine translation.

#1 Speed and Volume

When the architecture is adequate, MT can provide nearly instant translation of thousands of words. That means that a real-time conversation could almost be established between two people without delay, even if the languages are different.

In addition, the high speed makes it possible to handle large volumes of text, optimising time even for projects that, at first sight, might seem complex. Finally, some MT engines enable you to work with content management systems to organise and tag the materials. This will be very useful if you decide to translate into more languages later on.

#2 Reduced Costs

Most machine translation engines are free. Or at least they have a free option and advanced—paid—premium features. Thus, we can safely conclude that, in general, it enables you to cut down on costs. 

But, we talk about cutting costs and not removing them because you would still need a language partner to edit the content, in the most common scenario. In the most promising (but rare) case, you would still need an experienced professional to review the output for approval. This way, you can validate how good the translation machine is in a certain context and for a specific type of material without putting your brand’s reputation to the test.

#3 Language Options

Depending on the MT engine, you will have access to many languages to translate to or from. For example, Google offers more than 100 options from different language families. Others, such as DeepL, offer almost 30, but with additional features, such as glossary functionality or the selection of language variants.

In short, with MT, you can translate multilingually and with the same agility for each language. This is particularly useful if you are thinking of launching products or services globally and uniformly.

#OptimationalTip: MT allows you to translate into several languages, but it is not advisable to use it as a tool for multilingual SEO. While it can help you with translation in general, you will need a professional expert to not only edit the content but also to check that the strategy is working (keywords, link building, etc.)

How Good Is Machine Translation: The Cons

So far, so good. But, as with any serious analysis, it is time to look at the weaknesses of this tool. Only then you will have the full picture to make a better decision on your business’ linguistic needs.

#1 Omission of Intention

As we know, intention says a lot about a speech. We do not only speak with words but also with gestures and expressions, body disposition, and so on. And although the written text does not always make these details explicit, it often gives us hints on how to interpret the information correctly. For example, with the use of irony and exaggeration. MT does not know about intentions and could translate an expression literally, causing it to lose the effect it should transmit.

#2 Lack of Context

The main disadvantage that this type of software has—and that makes how good is machine translation a tricky question to answer—is context-free translation. MT focuses on translating words and phrases but cannot take into account the larger context in which the speech is embedded. This sometimes leads to reference errors, the ambiguity between designations, and more.

#3 Problems with Equivalence

With MT, the quality of the results is not fair and even for all language pairs. This is because there are some languages that are morphologically richer (especially with respect to inflexion); this poses a challenge when the translation is between two such dissimilar types. Here, a human language partner can develop strategies to compensate (if necessary) or clarify, but sadly MT is not up to the task.

#OptimationalTip: A fairly common problem with MT is the inaccurate translation of pronouns and forms of negation. Fortunately, an experienced language partner will notice this and make all the necessary adjustments.

Best Use Cases of Machine Translation

So, with both advantages and disadvantages, when is it really advisable to use automatic software? Or for what type of content? That is, without putting sensitive information at risk, compromising the company’s reputation or getting into hard-to-solve situations.

Here are some options, but first, a disclaimer. As there is no Google Translate for business type option to ensure the integrity of the results, you will need to undergo a review process. This is not a warning; it is just a word of advice. (We do not want to say, “We told you so”.)

  • Internal Communications and Emails: If speed is what you need, the machine can be your ally. Particularly for companies with multiple sites or global teams—in which several languages are spoken—and what matters most is that everyone receives the information almost at once.
  • Technical and Product Documents: Technical translation is an excellent match for MT, as it usually involves a great deal of text, but is divided into short, logical segments. The complexity is often lower than in other fields as it strives for readability and understanding.
  • Topical content: If you have content that is not consistently evergreen, has a less direct focus or has a short lifespan, MT can help you. Website reviews, product descriptions or trending posts on social media; just make your choice.

In these scenarios, we would no longer ask how good Google Translate is but how efficiently we could communicate the core message.

#OptimationalTip: It is better to stick to one hundred per cent human translation when branding plays a role on stage or with texts with artistic input, where creativity is important. Also, with durable assets and high-visibility content, such as blogs, client-oriented newsletters, press releases, website IU, etc.

MTPE: The Human Touch for Quality Assurance

If you want to ensure that your texts are clear, grammatically correct, and mistake-free, you cannot do without machine translation post-editing or MTPE. Better said, you will not be sure how good the machine translation is for you until having QA in place.

And this is simply because the human eye can do what the machine cannot: Go beyond words. Look for sense and aim for consistency in the tone and voice that can define a brand or reflect its values.

And, not least, prioritise between the words that best suit the goal you are pursuing with your text. We all know that give, offer, deliver, and provide can mean more or less the same thing, but are they really and exactly the same? It will depend on how you use them in each specific piece of content and for what purpose.

The MTPE covers two interrelated processes: Editing and proofreading. Here are the aspects linguists analyse during these instances to guarantee top-level results and fidelity to the source text.

#OptimationalTip: If you opt to combine machine translation with editing by an experienced CAT tool professional, you will benefit from the use of translation memories, which will provide consistency on translated segments that are the same or similar.

Aspects Under the Microscope During the MTPE

Editing and proofreading can be light (LPE) or full (FPE). This will depend not only on the quality of the machine translation output but also on your needs and expectations.

The good thing is that if you want to achieve uniform content with respect to other materials in your company, you can offer your language partner glossaries, terminology bases, and style guides to work with as guidelines.

In the review process, the professional linguist will look at the following aspects:

  • Grammatical accuracy
  • Proper order and syntax
  • Formatting and numbers
  • Consistency
  • Fluency
  • Specific or technical terminology
  • Writing style
  • Adherence to glossaries and brand manuals
  • Brand and product names
  • Multilingual SEO aspects (if content for websites)
  • Regional localisation
  • Words to be kept untranslated

And more! It all depends on how deep the professional linguist has to go and the extent of the text’s speciality.

#OptimationalTip: Remember that Google lowers the ranking for duplicate content. Unedited machine translation materials, rendered with any tool, fall into this category. Do not leave your SEO to the sole mercy of technology.

Last Thoughts about MT: In Professional Hands, a Tool for Expansion

With all this said about how good the translation machine is (or not), it remains to address it in a specific context: Yours. That is, how good it is for you, your business and your projects. For this, the first thing you should do is to classify the type of texts you want to translate, so you can determine whether they are of high or low visibility, internal communication, etc.

Then, consider the resources available, especially time. Also, how many languages you want to translate into, and how close they are to each other. After all, translating from Spanish into Portuguese is not the same as translating into Arabic or Mandarin. And last, but perhaps the most important step: Find a reliable linguistic team to associate with, one you can get your doubts out of the way and that can adapt to your strategy.

They will give you real advice on when automatic translation can be a good asset and offer you a sharp editing and quality assurance process.
Ready to start your multilingual strategy and benefit from state-of-the-art technology? Contact us today so we can guide you through your next steps.