Accuracy of Automated Language Translation

Last Updated February 27, 2020

Like every other industry sector, translation and language services are undergoing an upheaval.

With Machine Learning based applications disrupting traditional processes, language services are adopting innovative ways to enhance services and language solutions.

Translation software now spans hundreds of smart applications and online programs, each packed with the latest tech tool. There’s no contention that automated systems have an excellent capability to creep into data points. This often remain undiscovered by human intervention, or might be difficult to address manually.

Also, operationally, automated language translation makes it easier, faster, and more convenient to manage translation.

Key features of automated language translation:

Cost effective

Some online automated translation services are available for free and, even if there is a cost involved, it  can lead to a decent ROI if planned and executed professionally.

If machine output is sufficient and does not require human intervention, then there is no recurring cost of human labor. There are many customizable solutions that result in a positive ROI in the long run.

These can be in the form of on-premises machine translation engines and services. That includes online SaaS models with a pay-as-you-use feature. That’s especially beneficial in case of GDPR and Data Privacy concerns. But it is essential to plan the volume, quality, timelines and domain.

It can do more

Needless to say, the core idea of automation is to extract maximum output, within minimum time. And automated language translation software does just that. They are easy-to-use and can do more and at a great speed.

Most automated translation programs come with in-built multilingual capabilities. They can translate hundreds of languages in seconds, and sometimes, at the same time.  A single solution may not provide you with the same level of accuracy for all languages.

Some solutions are more effective for certain language pairs whereas some are effective for the domain.

But every pro comes with its respective drawback, and automation is no different. While it is indeed bliss for many to harness the benefits of automation, it might not be the optimal solution for all services in the language and translation industry. Nor can it be completely failsafe.

Automated Language Translation – Walking the Tightrope

 Linguistics is a vast, enigmatic space that involves thousands of years of understanding of culture, emotions that drive that culture, and spoken and written words that have evolved as a byproduct of such human emotions, interactions, and actions. When you translate, you often can’t replicate the original emotion.

However, with a human interpreter/translator involved, they would know the finer nuances of both the origin and destination language and will not translate every word or phrase verbatim. The idea of language translation is to retain the emotions and sentiment, and underlying cultural references of both.

Machine learning, as we know it, is still an evolving area and has not reached its peak potential. Hence, it still remains a challenge when you use machine intervention for deep learning or explore emotional intelligence.

Automatic translation isn’t at a  place where it can be as accurate as someone speaking as their natural course of learning and growth within a certain cultural environment.

Context is King

When dealing with languages, the most important element to look at is context, and not always, content. Certain cultural references used in languages might not be relevant in another language, even though there is a logical word/phrase for it.

For instance, the metaphor ‘hit the ball out of the park’ is a classic baseball reference that applies to American English and to some extent, British English with respect to cricket. But in a country, where neither of these sports is popular, it can lose its relevance in the local language, since there is no context of such a phrase in the regional culture.

As such, the automated language translation is still not able to achieve the creative aspect of linguistics like puns, metaphors, adages, or slang, which give a character to the content that is being translated. A human translator, on the other hand, can fill in these gaps and make it completely contextual.

Language and terminology can also have the same effect in the case of technical translations. For example a word as simple as “Home” has different translation depending on the context or the word “Fan” can have different translation when used in manufacturing.

This takes us to the next element of translation- grammatical correctness. The grammar of a language comes from its structure. Languages like Spanish and Italian when machine translated into English result in a far superior output as compared to German.

This is because the grammatical structure and the semantics of these two language pair are very different.

Automation has not yet achieved the accuracy of the local lingual phrasebook or the experience and skill that a human translator brings. Using automation in such cases can leave the reader/listener at a loss.

The Way Forward

Since we are in a time and space of constant tech innovation, we cannot escape integrating automation in every business service and function. That is why we need to strike a balance, where automated language translation can be supported by human skills.

With applications like Neural Machine Translation (NMT), it might be the way out. The program deploys a neural network-based technology that uses certain algorithms to break down linguistic nuances and cultural conventions, and then it is followed by human post-editing. By combining the best of both, this process yields better outputs.

To summarize, automated language translation has to be well balanced with the right intervention of human intelligence to deliver the right results

Related Posts

Summa Linguae uses cookies to allow us to better understand how the site is used. By continuing to use this site, you consent to this policy.

Learn More