– Language translation is a widely adopted AI use case, with companies like Google making advancements in the field
– Google Cloud’s Translation API introduces a new generative model that allows for more flexible and real-time translation
– Adaptive Translation, integrated with Translation LLM, has shown significant quality improvements, making it a valuable tool for companies like Smartling for localization and content management.
Among AI use cases, language translation is widely adopted by companies like Canva and Bloomberg to make content more accessible. Google has introduced the transformer architecture, powering large language models and making advancements in AI Translation. The Translation API now offers a new generative model, the Translation LLM, for longer text translations, while NMT is more suitable for short text or chat conversations.
The Adaptive Translation feature, integrated with the Translation LLM, allows customers to import a small dataset of translated examples to optimize translation outputs in real-time. Smartling presented at Google Cloud Next ’24 on responsive translation using generative AI, showing benchmarks of Google Adaptive Translation outperforming Google Translate by up to 23% in quality. The VP of AI at Smartling, Olga Beregovaya, praised Google Adaptive Translation for its customizable and cost-effective solution for clients with sparse data scenarios.
The Translation LLM can be tested on Vertex AI Studio in translation mode, offering more options for businesses to accelerate their translation use cases with AI. The Adaptive Translation feature enhances the translation process by providing refined context and improving translation outputs to better fit style requirements. Overall, the integration of generative AI in translation technology is making translation more flexible, efficient, and accurate for companies and organizations.