Machine translation (MT) has been getting a lot of attention recently with the introduction of neural MT engines. No doubt this is an important development for the translation industry, but there are still many questions surrounding MT and its suitability for various translation needs.
- Do these tools work equally well across all languages, such as those spoken in Europe, Asia and the Middle East?
- How well does MT translate creative, technical or eCommerce content?
- Can MT be used effectively in a professional setting for business content?
We’ll get to the answers shortly but as you’ll see, the use of MT depends on language requirements and what type and volume of content you have. There are many MT options available; from open source to proprietary vendors. Some are free and others are costly. An experienced translation agency like Acclaro can help you make an informed decision and implement the MT approach that works best for you. You might find that MT on some (or all) of your content is a worthwhile investment, allowing you to realize cost savings and decrease your time to market.
Depending on the type of MT used, you will have to weigh cost, speed and quality. Your choice will also depend on what content types you like to have translated. You will also be asked: “What are the quality expectations for the different types of content you may have?”
To figure out what’s best for you, let’s start by understanding the different types of MT used today. As you’ll see, some technologies are old and some are new, but each has a specific place in the translation landscape.
4 types of machine translation and how they work
Machine translation was first used in the 1950s. Since then, multiple developments have propelled the technology forward to help get better quality translations. Types of MT include:
- Rule-based machine translation — a collection of “rules” governing the use of language developed by linguists is used to translate to the target language. For example rules around the use of formal and informal version of “you” in German. Although this type of MT was the first commercially available, the technology is no longer used as such.
- Statistical machine translation — systems use algorithms to produce text that appears to be the best selected from millions of possible permutations. In a hybrid system, rule-based and statistical translation are combined. Like rule based MT, statistical MT is no longer used due to effort needed to maintain the system and low translation quality. However, both systems have been blended into hybrid systems to improve translation quality.
- Neural machine translation — The latest breakthrough came in 2016 when Google introduced neural machine translation. This form of MT uses a neural network or Artificial Intelligence (AI) modeled on the human brain to predict the most likely sequence of words.
- Adaptive machine translation — Another innovative development came in 2016 when Adaptive MT was developed. Translators work with MT suggestions interactively while the MT engine is trained in real time. An MT engine learns new terms and phrases in the right context and tone of your business for higher quality translations. MT platform providers can further optimize MT engines for speed, quality, and cost for large-scale localization projects.
Is MT a good option for you and your content?
MT, when used in the right application, is usually faster than human translation. This typically creates efficiencies that can lead to faster deliveries and lower costs. The key is to realize that not every piece of content is best suited for MT. Sometimes, a human translator is going to produce the best result. Other times MT and a human translator can work in concert to deliver exactly what you need.
While machine translation can handle many types of content, there’s a lot to process in figuring out how it can work for you. Each type of machine translation has pros and cons. One challenge with MT adoption has been a binary use/don’t use equation. But as it’s integrated seamlessly into translation tools and processes, it will become a real-time assistant to the professional translator and accelerator of high quality human translations.
Let’s look at some examples to make this clear.
The type of content you need translated will ultimately drive the decision on whether to use MT or not. A good rule of thumb is that the higher the value of the content, the more likely a human translator is the best fit for the job.
MT for FAQs and other knowledge base content
Great candidates for using machine translation to assist human translators are customer support content, eCommerce content like product descriptions, FAQs and technical documentation, where understanding is the most important goal of translation.
Rather than spending the money on a process that exclusively uses professional translators (as you might for your corporate web content), you can realize cost savings for this type of content by integrating MT technology. FAQs and product descriptions are written in an informative manner without the nuances of more creative content. Also, the content generally doesn’t have the same level of visibility or criticality as marketing content, so the objective is understandability rather than perfection.
Often, human edited or even “raw” MT is used for content that wouldn’t be translated otherwise due to budget and time constraints. We typically recommend this to clients that have well-structured, informational content that consists of millions of words.
MT for legal and similar content where accuracy is needed
Moving up the content hierarchy are product packaging, apps, a user dashboard and legal. Accuracy is extremely important here and having a human translator do the translation is the obvious choice. However, adaptive MT integrated into Computer Aided Translation (CAT or TM) tools can also play a role in the initial stage of the process, possibly making it easier and faster for the translator.
As the need for accuracy rises, the precision of a human translator becomes even more important, as you’ll see in these next two levels of translation.
Getting the click with creative and conversation marketing
When you are trying to gain a conversion — via marketing content, website, emails, sales presentations, videos, etc., you need a high level of accuracy and want a human translator to ensure this type of content is translated correctly.
You’ll also want a human translator to do the bulk of the work, if not all, for highly creative material where your customer starts his or her journey: SEO, pay-per-click ads, landing pages, etc.
The future of machine translation
In the 1960s, the available systems were inaccurate and costly when compared to human translations. Today, despite limitations on language and the inevitable potential for errors, machine translation is quickly becoming an important part of the professional translation industry.
With this growing innovation comes timeliness. Clients can receive drafts on their translations more quickly than ever before, speeding up the time-to-market.
The technology is awe-inspiring and will only push further ahead. Developments like neural machine translation by Google and Microsoft are making machines vastly more useful for businesses.
Neural MT can understand the similarities of words, consider entire sentences and “learn” complex relationships between languages. Fluency (proper grammar and readability) has come a long way in neural MT, accuracy is better as well, but still requires thorough post-editing. A human translator now has the ability to work in congruence with a machine to start the translation process, and then fine tune for a finished product.
Neural MT is not without some drawbacks, though. Language combinations with strong neural engines are limited. In the coming years, more languages will come onboard and the translation process will be further streamlined.
We work closely with clients to assess where MT may make sense for them, and have experience with a variety of machine translation platforms. Since we are “platform neutral”, we weigh all of the options to select the right type of MT approach for our clients to fit their goals, needs, budget and content.
How we assess if MT is a viable option for you:
- Requirements Analysis – in this first step, we evaluate your content volume and type(s), language combinations, quality expectations and privacy requirements.
- Content Analysis – we assess if your source content is suitable for MT from a technical and linguistic perspective. This includes file types, structure and formatting.
- MT Platform Evaluation – based on the first two steps, we evaluate what MT platform is best suited for your localization project. This may include an MT output evaluation where we apply machine translations from several platforms and review and evaluate the output.
If you’re considering machine translation and would like to find out what options are available to you, contact us today for an evaluation.