The globalized nature of business today encourages organizations to make services, or at least their presence, known to multiple markets across multiple languages.
The development of generative AI is helping increase speed and efficiency of translation services, leveraging algorithms to auto select the most appropriate translation engine based on content type.
Smartling Translate, released last month, leverages previous assets such as a glossary and translation memory through its LanguageAI proprietary technology, and lets users copy and paste text or drag and drop files up to 200MB. The platform also implements style guidelines and can be trained to follow the brand voice, as well as locale-specific conventions, grammatically accurate terminology handling and the proper use of linguistic gender preferences in translations.
Smartling CEO Bryan Murphy explains including brand voice in translation is important because it helps maintain consistency and coherence in communication across different languages and cultures.
“A brand’s voice reflects its unique personality, values, and style of communication, which are carefully crafted to resonate with a specific target audience,” he says.
From his perspective, it is essential to have brand consistency in translated content to improve communication effectiveness and create a consistent, emotional connection that gives the company a competitive edge.
“The way you do this in translation is by utilizing tools such as inclusion of your glossary, the specific brand terms and phrases you utilize in marketing; translation memory,” he says. “This is essentially a database of previously translated words and phrases so that they are reused consistently with your brand every time and in every language.”
Smartling’s platform performs all its work in a private cloud environment, which Murphy explains offers additional security because many public translation tools have rights to harvest your data, which is bad for privacy and security.
“Maintaining a safe and secure environment for translation that protects confidentiality is essential and safeguards data, respects intellectual property rights, builds client trust, ensures compliance and mitigates the risks associated with unauthorized access or data breaches,” he says. “It is crucial for both the protection of sensitive information and the reputation of companies that translate.”
Jayaprakash Nair, Altimetrik’s head of analytics, data science, machine learning, AI and visualization, explains it takes time and effort for humans to pick up a new language, let alone become proficient enough to play the role of a translator.
“One human cannot pass on all this hard-earned knowledge in an easy manner to another human,” he says. “Machines, on the other hand, can replicate their knowledge of languages easily from machine to machine.”
He notes that until recently, humans were better than machines in the contextual depth of conversations.
With Large Language Models (LLMs) trained on billions of parameters, even this distinction is becoming thinner. Generative AI technologies can learn, converse and translate in various languages, and this trend is only expected to improve further.
“Document translations are making it easier for us to understand text written in foreign languages,” he adds. “Real-time AI translation enables foreigners to have meaningful conversations with each other.”
He says organizations are known to have a “voice” or “tone” to their branding campaigns–some are brash and explicit, some are subtle, yet others are polished.
“The AI-engine needs to be trained and tuned to the particular voice of the brand, as well as its terminology, instead of using the off-the-shelf AI services as-is,” he says. “This can be done by training the pre-trained LLMs using organization-specific data and appropriate prompt engineering.”
Nair says there are several such examples where AI based translation is impacting the transaction services market, and while there are several kinks yet to be ironed out, from his perspective it is making steady progress.
“AI based engines are known to stumble in nuanced language situations like satire or sarcasm,” he says. “So, in a batch translation setting, it is preferable to have the AI-generated translation quality-checked by a human.”
He adds that since those flaws apply to a real-time conversational setting as well, it is prudent that AI-based translation services are used very cautiously in sensitive discussions and negotiations.
“They should only be used in the presence of an expert human translator,” he advises.
With the combination of humans and technology, much can be accomplished.