The Pros and Cons of Machine Translation

Let’s get straight to the point: I cannot think of a single linguistic topic that is the object of more scorn and mocking than machine translation. And it’s not just linguists who raise their eyebrows at the mere mention of the term. Yet, like it or not, machine translation is now a reality for the industry, and for the general public with free online platforms like Google Translate. In this context, you can’t help but wonder why, then, machine translation gets so much bad press, yet is gaining such momentum. Let’s take a look at some of the pros and cons that may contribute to these conflicting attitudes.




Savings

The first benefit of machine translation, and probably one of the main reasons why people opt to use the technology over professional linguists, is savings. If all you’re looking for is the general meaning of a piece of content, for internal communication or online chats and forums, for instance, then a machine translation may well suffice.


In fact, you could argue that, with the increasing prominence of machine translation, the general public has come to accept and expect it, warts and all, whether it be to translate a post on Facebook or get the gist of the customer reviews on a product you’re considering buying. As long as it is clearly specified that a machine translation tool has been used, then any inaccurate or literal translation won’t necessarily impact negatively on your brand image or customers’ trust in you. The context, or score in the case of reviews, will likely make up for any deficiencies in the translation output.



Speed

In our “click and get” world, we are more and more accustomed to obtaining things at the click of a finger or mouse. The same is true of translations, especially for content that is quickly consumed and forgotten. To take the example of customer reviews again, it’s highly likely that their volume and turnaround for all of the products sold by a given company would simply be too high to be managed by human translators, and the business advantage of a pitch perfect translation over machine translation output would not be enough to make up for the loss of time, effort and budget this would entail.


It is, however, essential to bear in mind that this compromise over quality is acceptable only when dealing with user-generated content, which the relevant company is not accountable for and is not likely to suffer from. For other types of content, post-editing is necessary to ensure the final output is accurate and understandable. Whether a binding legal document, a product description or an instructions manual, it’s not hard to see the potentially damaging effects of poor machine translation output without any form of post-editing.


Don’t be fooled: Even content that you believe to be straightforward may not always fare well with machine translation: I remember one website promoting rental properties that had been translated using a machine translation engine with no post-editing which had turned the English “designer flat” into a French “creative dish” (“flat” and “dish” are the same word in French). How do you think customers will view your business if you’re unable to use the right product name in your virtual shop window?


And that’s where things become a little blurry. How much post-editing is required to avoid any risk of misunderstanding or harm to a brand, and, in light of this, is machine translation still worth it? After all, the more time and effort you need to spend post-editing content, the less speed advantage you gain in using machine translation.




The post-editing factor

The amount of post-editing required will largely depend on the type of content and language used. Let’s consider the case of a legal contract: Such documents typically employ highly formulaic, even repetitive clauses, and it is likely that a machine translation engine would have enough data to output a fairly accurate text. Post-editing should then allow you to smooth over any inaccuracies or inconsistencies while still gaining in speed, at least if you don’t have a good translation memory at hand.


If, however, you do have a good translation memory, you may find yourself with a high number of 100% matches and it may prove just as quick to amend the high fuzzy matches as it would to post-edit the machine translation output. You may even benefit from greater consistency in terms of the brand’s terminology, style guide and tone of voice. Templates are another way to increase efficiency and speed. This is particularly true when using placeholders, allowing you to adapt a common text according to different data (opening days and times, prices, brand and product names etc).


Needless to say, the more creative and unique a piece of content is, the more post-editing will be required, to the point where machine translation is no longer a worthwhile option. Pretty much any type of content that requires localisation or transcreation, such as white papers, press releases, campaigns, taglines and slogans, is better off being taken care of by a human translator. Other requirements, such as character limits, the use of specific keywords or hashtags, may also make machine translation a no-go.




So, what’s the conclusion? Well, with the constant evolution of machine translation, there is little doubt that it will continue to gain in prominence. Yet, the technology still has a long way to go before it is able to successfully translate turns of phrase, puns, nuances and cultural references. When choosing between machine translation and human translation, it is therefore essential that you assess the quality and level of post-editing that you require, depending on the type of content you wish to translate.



22 views0 comments

Recent Posts

See All