Should I use machine translation for my content? is one of the most asked questions we get from people contacting us for the first time. In the past, it never seemed that machine translation was good enough to be what many people believed it would be – a replacement for human translation.
However, humans have been fascinated by the possibility of creating artificial intelligence which could go toe to toe with humans for a very long time. It just seemed that we will never get there. What’s the situation now and should you use MT? Today, we will try to answer that question.
Machine translation improvements
One fact that needs to be established is that without a doubt we have made great advancements in machine translation. Since previously used statistical machine engine models (using a huge set of data to guess the best translation based on statistics) were limited, a couple of years ago many of the leading machine translation providers (like Google Translate) switched to neural machine translation.
Although neural machine translation still uses huge sets of data to draw from, it now tries to actually understand the whole sentence before translating it. The results can be seen, especially if you use MT for the most popular languages, such as English, German or Spanish.
What does this mean for humans, and can we now replace them? We are not there yet, but there are some benefits and some drawbacks to using machine translation engines for the localization of digital content.
The benefits of MT
The benefits are, clearly, speed and cost. Machine translation is a lot cheaper than human translators (although not entirely free). Moreover, it can also be implemented quickly and produce instant results, so it’s useful for translating content which needs to be readily available to users (like support or chatbots). For companies looking to translate huge amounts of data, these arguments are enough to make them use MT.
One needs to remember though that like with most services, clients will tend to ask for the translation to be cheap, fast, and high-quality. MT can help with two of the aspects, while the third one is often debated, but can be listed as one of the main cons against machine translation.
The birth of machine translation and human review approach
To bridge this gap between pros and cons, humans have called upon, well, humans, to improve the machine-translated content. This led to the emergence of machine-translation and post-editing.
Essentially, this means that a machine translation engine will instantly and at low cost produce an ok-ish translation, which native human speakers will then review. It is a good trade-off, as reviewing a translation is cheaper than translating the content from scratch, but there are also limitations to this approach.
The type of content question
If you translate a repetitive, simple text, such as a list of products, machine translation, and post-editing will make a lot of sense, especially if you are using a Translation Management System which can help the reviewer to quickly change a repetition only once, since it will be then propagated throughout the translation automatically by the TMS.
Machine translation has also started to influence the content in the sense that many companies have realized that they can simplify the language used in most of their content. Like this, they allow for greater use of machine translation and reducing the cost of localization in general.
If you have to translate marketing content with a lot of idioms and complex linguistic constructions, then machine translation will just create a mess which the reviewer might not be able to elevate to a level that will satisfy your needs. In short, if you need to localize and transcreate, then machine translation might not be the best option.
Should you use MT? (or TL;DR option)
To help you answer that question with a definitive yes or no, we created a bullet point list that you can go over:
- Machine translation can be an option if you need a fast and cheap translation
- Machine translation works best with the most popular languages, such as English, German, and Spanish
- Great for translating content which is not complex linguistically
- Should not be used for the content which needs to be transcreated or heavily localized
- Major areas for implementation include (most of the) content on websites, product lists, e-commerce shops, simple manuals, and chatbots
- Can be improved further with native human review
Is this the way forward for you?
In any case, it would be best for you to reach out to people who have experience in this, and who can estimate whether your content can be handled using MT, with or without human review.
If you want to talk about this, reach out to us, and we will talk about our experiences and propose the best approach for the translation of your content. As always, there are no strings attached, so why not check what can be done?