2023 looks to be the year when machine translation becomes an integral fixture of the content marketing landscape. This prediction is supported by top-line data points in Technavio’s in-depth survey of the content marketing industry released in August 2022.

The report anticipates substantial revenue growth at a galloping 15.8% through 2026, accounting for a massive $487 billion income differential over that 5-year stretch.

Most remarkably, the report projects that a whopping 38% of this growth will come from the APAC market. For the vast majority of content marketing, that massive chunk of change has, until now, been mostly off the table. Which marketers, after all, have the luxury of focusing on reaching a mindshare of the sprawling multilingual markets of Asia and the Pacific, spanning 48 countries? Think China, India, Indonesia, Japan, Thailand, and Vietnam. That’s a total of 4.3 billion people—about half of humanity—of which only a tiny fraction speak English. 

Sorry, Australia and New Zealand. Sheep don’t count.

For content marketers who don’t want to ignore those massive audiences and leave that revenue behind, the question becomes: how can we economically translate our marketable content in a scalable way to these almost exclusively foreign-speaking markets? Until now, content marketers going after these massive but hard-to-access targets have had to staff up internally or hire agencies, country by country, for translation, website localization, and language-specific marketing. That’s expensive and time-consuming in addition to being difficult (if not impossible) to manage at scale.

However, in 2023, machine translation is fast becoming a viable option for massive and automated content translation. What changed? The answer is NMT: neural machine translation. After decades of R&D, NMT is finally reaching a point of market-readiness that is revealing to be a game-changer in several ways. Here are some take-aways about leveraging this fast-maturing technology for the practical tasks of content marketing.

 How does NMT supersede previous iterations of machine translation?

Neural machine translation (NMT) is based on a neural network that uses artificial intelligence to mimic the messaging of the human brain.These neural networks are made up of neurons and synapses that enable high-end computers to catalog and map any given language and accurately convert it to another. 

Each network applies a variety of learning processes, which resemble to some extent the way human beings learn and think. NMT systems are trained using vast quantities of data, aggregating millions of vetted translations completed by human translators, then interpret from that multiplicity of samples the rules of how language is constructed and expressed. 

The GIGO rule of computing—Garbage In, Garbage Out—applies here. One reason why early NMT translations were criticized as dull and lifeless was that the samples used for machine training were often derived from UN and EU parliament documents. 

The samples being used for training have vastly expanded in recent years. Apple began using samples derived from the language iPhone users texted. Meta has been doing the same for social media posts on Facebook. 

Unlike SMT, NMT does not work word-for-word but sentence-by-sentence, or even paragraph by paragraph, evaluating all possible translations of a text and then choosing the optimal way to recreate the intended and contextual meaning. The result is a translation more accurate and natural-sounding than was possible before.

Can free NMT be leveraged for content marketing?

For SMEs, content creators, and marketers, it boils down to whether they will try to get by using free or cheap translation software solutions. The alternative is to hire agencies that specialize in NMT and know how to leverage it in translation projects.

There’s a trio of capable performers that use NMT: Google Translate (supporting 133 languages), Microsoft Translator, and the smaller but equally capable DeepL. Content marketers should check out each of these, just to be familiar with what’s out there. The basic features for all three are free to use, though API applications of the programs for automated translations have a payment wall. 

But how do we measure their effectiveness side-by-side? Enter NMT comparison tools. Unlike using conventional NMT engines, content creators can use these tools to assess the translation quality across different services by displaying side-by-side translations from multiple engines. One of these NMT comparison tools is Machinetranslation.com, where content creators can simultaneously use various NMT engines, like those mentioned above, and get an automated analysis of the results generated from the NMT engines. Most recently, it has launched a Beta feature, Context Translation, that allows users to generate more personalized machine translation content. 

However, even though NMTs have greatly improved over the years, there still remains a quality gap between the top NMTs and the best human translators. In 2020, the estimated accuracy of machine translations, according to linguistic experts, ranged from 60%-90%, though it’s estimated that NMT improves 3-7% annually. So, by 2023, we should expect NMT accuracy to be somewhere in the range of 72 to 98%.

Where NMTs excel even now, however, can be of keen interest to content marketers seeking to expand markets at a low cost. These use-cases involve documents or software with a limited set of predictable words and jargon. 

Technical documentation: user manuals, operating guides, instructions
Online commerce content, especially catalogs and product descriptions
Business software localization for computers and phones
Marketing and financial reports
Standard business websites

What this list lacks, of course, is anything creative, imaginative, or persuasive. NMTs do not excel with content intended to play on human emotions. They are better at “cold” and “dry” translations. Don’t entrust them with your branding or marketing campaigns. 

What value can NMT-capable translation agencies bring to content marketing?

Even for the coldest and driest of documents, you still can’t get by on NMT alone. Always be sure to have a native-speaking translator review each document in each language before it goes out the door. They will be able to provide quality assurance and constructive suggestions for anything translated by an algorithm or another human.

If you are working with an external language service provider (LSP), who are experts in the translation, interpretation, and localization industry, you may want to assign someone internally or take it upon yourself to learn the professional translation process. That way, your team can eventually take over these post-editing tasks.

Learn to do it yourself. Leverage a hired LSP’s consulting smarts and accumulated know-how to get started. This will allow you to gradually acquire independent knowledge on how translation and localization is done. Apply software at the API level for content you need to translate in volume.
Over the course of months and years, you can bring some of the tasks initially performed by an LSP in-house to save your organization time and money. But keep the translation pros in the loop, for QA and consulting as well as help translating documents and software that require a creative flair and emotional impact.

Author bio

New York-born Richard Koret is a full-time writer and editor, with a degree from Princeton and three decades of experience in content creation, marketing communications and online publishing. He is a serial entrepreneur, and served as COO and CTO of Africana.com, sold to AOL Time Warner for 8 figures. He is a world-wandering nomad currently residing in Thailand.