in All Things Localization

In the world of localization, translation accuracy is often a critical concern for clients and translators alike. The traditional model of localization has several widely-accepted QA models which are used to guarantee accuracy such as, ASTM F2575-06, EN 15038, the LISA QA Model, SAE J2450, the ATA framework, and TAUS DQF. When it comes to finding fast, accurate linguistic solutions that can meet the demands of enterprises, such as crowdsourced translations, these quality standards become vastly more difficult to target. Nevertheless, many enterprises view the crowdsourced translations model for sourcing global content very favorably.

Few industry experts can forget when, in 2016, Google Translate announced that its translation accuracy had improved to an average of 4.3, only .3 below the accuracy level of a human translator. This drastic improvement in accuracy was due to its reliance on artificial intelligence which allowed it to utilize human translators through a crowdsourcing model which updated the machine’s database, thereby improving translation quality.

Moving towards a model of crowdsourced translation means that the small well-trained pool of qualified, professional translators is suddenly flooded with a flock of bilingual individuals. Quality that was once taken for granted by businesses and localization companies alike can no longer be assumed, nor do previous models of quality control systems apply to the new model of localization output. New methods of accuracy control must be put in place to utilize the knowledge of the crowd while maintaining the wisdom of language and cultural experts.

Testing your translators is key to maintaining quality translations; it is important, after all, to know that your team can perform well for you. One method to test for quality is to compare a translator’s work to known translations (known as gold answers). A second method of testing is to deliver one task to multiple individuals to identify how often they agree with one another.

Related:  Hyperlocality in the Age of WhatsApp

Training and managing translators correctly is also key to maintaining accuracy. By utilizing a professional linguist amongst amateurs, you can expand on their knowledge and train the bilinguals to understand the intricacy of translation. In order to understand which linguists are experienced and which are not, standardized tests may be given out to determine the skill level. More qualified translators can then coach and provide feedback on the quality of an individual’s translation, thereby building up their knowledge and improving the linguistic output.

Necessary to all translation processes, post-editing is the key piece in maintaining quality in your translation. While linguists may be tested and evaluated beforehand, the true accuracy of a piece can only be determined after translation has occurred. It is during post-translation that an individual can finally fully understand the context of the translation. During this time, true localization happens. Small cultural idiosyncrasies can be observed and adjusted to the targeted locale and discussion of glossary terminology can ensure translation quality across content.

Crowdsourcing translation offers many benefits to both linguists and companies alike. Yet, like all localization projects, it is important to implement careful strategies for maintaining accuracy throughout the process. In many industries, a small translation error can have a huge impact. Be it a life or death situation, your company’s reputation, or the difference between success and failure in a local market, it is always important to work with a language services provider who understands your content, industry, and brand strategy.

Learn more about CSOFT’s quality guarantee.

Written by Megan Robinson, PR Manager at CSOFT.