With the world changing rapidly in the face of technological revolution, language workers are understandably trepidatious. But is it warranted? While any change — particularly rapid, big change — is scary, there may be more cause for optimism than one might think. 

Certainly, that’s the prevailing attitude at Crowdin. They envision a future where emerging technologies like AI make language work easier and better rather than replacing workers or flooding the field with low-quality content. Remember the movie Her starring Joaquin Phoenix? It’s entirely possible that’s not far from the future that awaits us: an AI assistant who boosts our careers and opens opportunities, all with a personal, startlingly human-like touch. 

In the language space, Crowdin envisions something similar for language work. They believe the language service provider of the future will have not just one but several AI assistants, one of which will likely be dedicated to translation tasks. These AI assistants will be bespoke solutions for each company, trained to capture its unique voice and serve clients according to longstanding preferences.    

“In the world of localization, AI is a hot topic — some wonder if it really works,” Crowdin Product Marketing Manager Diana Voroniak said. “At Crowdin, we’re sure it can make a big difference, but it’s like driving a car. If you sit there and it doesn’t go anywhere, it’s not the car’s fault; it’s about your driving skills. Same with AI — you need to learn how to use it.” 

Fortunately, Crowdin’s AI Translator makes learning easier than ever through the steady process of fine-tuning. The AI assistant can turn your preferred stock large language model (LLM) into something startlingly close to a bespoke solution through exposure to your existing translation memories (TM) and glossaries. And thanks to the system’s flexibility, you can do it at your own pace. Simply initiate the fine-tuning process anytime your TM or glossary gets a significant update. In the not-too-distant future, this process will likely occur in real-time. 

It’s yet another exciting addition to Crowdin’s proven translation solution. And that solution is based on one vital ingredient: context.   

It’s no secret that context is vital in producing quality translations. Machines can understand language perfectly, but because words have multiple meanings and permutations with rules and exceptions shifting from language to language, it’s all too easy to produce a bad translation with insufficient information. Humans, on the other hand, have the intuition to solve that problem, so why not lean into our unique talents? 

Crowdin’s system allows for unparalleled specificity in establishing required context for better quality right from the jump. The secret? A layered approach that feeds the machine system context at the project, file, key/string, and translator session levels. 

It all begins with the project-level context, or prompt, which should typically include consumer, product, company, and business domain information. Localization managers modify it with project-specific details, and Crowdin then fills in necessary context like translation memory matches, glossary terms, and approved translations before prompt execution.

Next, the file context brings greater specificity and flexibility to individual files in your Crowdin project. Think articles and Word documents that require translation — Crowdin helps the machine understand the nature of both the file and the required translation. 

But what about the translation of more visual elements? Software UI elements — which typically use short texts that would challenge man and machine alike without context — are essential translations, but how can the system comprehend what’s required? Enter the key/string level context, which establishes a roadmap for that exact task. It’s also great for text- and graphic-heavy projects like website translation. 

Finally, the session level gives linguists, rather than project managers, a portal to their preferred LLM for assistance on demand, and once again, the more context provided, the better. 

The modern world brings complicated challenges for language and localization work, particularly at the speed, quality, and price point clients expect. And the wide range of translation projects, from websites to software to articles and documentation to marketing copy, complicates the situation yet further. Machines speed that process along, but they still need a helping hand. And Crowdin is that helping hand, an all-in-one system that provides the method to routine translation madness.    

“We’re excited for everyone to test and improve AI translation tech with us,” said Diana. “Let’s learn and share our thoughts at https://community.crowdin.com/.”