Intento, Inc.

🎉 We’ve published our 8th annual State of Machine Translation report. It analyzes 52 MT engines and LLMs across 11 language pairs and 9 content domains. Get your copy here ➡️ https://hubs.la/Q02zD6Z70 Here’re some key findings: 🎯 LLMs reshaped the MT landscape and now account for 55% of top-performing models 🎯 Quality varies by language and domain, with LLMs excelling in Colloquial, Education, and Entertainment 🎯 Traditional Machine Translation (MT) still outperforms in specific language pairs and domains 🎯 LLMs are less expensive but slower than MT engines 🎯 Open-source LLMs lag behind commercial offerings 🎯 Customization through translation memories, glossaries, and prompt engineering is key to eliminating errors Explore the full report to learn more about the evolving capabilities and potential of LLMs. e2f, inc.

Patronus AI released a hallucination detection LLM called Lynx. It is built on Llama-3 and does well on public hallucination evaluation benchmarks, outperforming GPT-4o and Claude-3-Sonnet on some of them. Lynx can be a useful tool to mitigate hallucinations in LLMs; however, it requires careful evaluation on your tasks. https://hubs.la/Q02GMq3b0

You can now import your Translation Memory (TMX) files into our Translation Storage to leverage it in all Intento plugins and integrations, including real-time translation use cases like live chats, on-the-fly website translation, and more. Book a demo to see how it works! ➡️ https://hubs.la/Q02FmrzJ0 While we initially developed Translation Storage to review and reuse past machine translations, it now serves as a data bridge between content and experience translation.

Last week, many sources reported that Google Translate now supports 110 new languages (https://hubs.la/Q02F0frH0). The whole list is here: https://hubs.la/Q02F05YJ0. Our systems automatically monitor API changes, but we didn't receive any advance notice. The reason is apparent if you read the original post carefully—these languages were added to the Google Translate web and mobile application, but they aren't powered by Google NMT; instead, they come from PaLM 2, one of Google's LLMs. This raises an interesting point in the debate about LLMs' ability to handle both popular and low-resource languages. They may excel at both ends of the spectrum but struggle with languages in between. For widely-spoken languages, LLMs outperform NMT by using more data. With low-resource languages, where NMT lacks sufficient training data, LLMs produce decent results due to their highly multilingual nature. For less popular languages and domains, achieving substantial quality requires dedicated evaluation and data curation efforts, which we may not see immediately. See our “The State of Machine Translation 2024” report for insights: https://hubs.la/Q02F06rr0

Anthropic has just announced its latest model, Claude 3.5 Sonnet, the first release in the forthcoming Claude 3.5 model family. While 2x faster and 5x cheaper than the previous flagship model, Claude 3 Opus, Claude 3.5 Sonnet outperforms it on a wide range of public benchmarks. Claude 3.5 Sonnet is immediately available for machine translation across all Intento connectors and integrations! Try in on your tasks today! https://hubs.la/Q02CHHqP0

Our latest blog post explores how Retrieval Augmented Generation (RAG) with LLMs can significantly improve glossary term translation ➡️ https://hubs.la/Q02BzQnD0. By retrieving relevant glossary entries and using them to guide translation output, RAG delivers more accurate and consistent terminology than standard neural machine translation. This cutting-edge approach combines the best of rule-based and neural methods for a powerful upgrade to glossary integration. Read the full post for details on our experiments and results!

💥 Join us today for a webinar with Vann Maxson, Head of Solutions at Intento, Inc. Vann will share insights on adapting to linguistic needs, improving source quality for faster localization, and integrating non-language teams into workflows. Register now! ➡️ https://hubs.la/Q02Bx2Rg0 XTM International

Hey memoQfest attendees! 🚨 Join Ben Warren from Amadeus and Steven Hayward from Intento, Inc. on June 13 at 12 pm to learn how Amadeus unlocked real value with GenAI in their memoQ workflows. They'll share the approach and talk about the real business impact that they see when adopting this technology in production. See you at this insightful session!

I was thrilled to meet in person Emil D. Atanassov and Riccardo Ermanno Natoli to learn how ServiceNow approaches UX localization testing. Another forward-looking company that focuses on localizing user experience, not just linguistic quality - and provides a great localization framework for partners like Intento, Inc. to support the same for the content their customers create. #Locworld51

Join us on June 12th for a webinar featuring John Weisgerber, Senior Solutions Engineer at XTM International. With over 20 years of experience in the industry, John brings extensive expertise to developing translation management technology in XTM Cloud. You’ll learn how combining XTM Cloud with Intento Enterprise Language Hub can boost source quality, reduce manual tasks, and speed up post-editing without extra resources. Register now ➡️ https://hubs.la/Q02zT7VY0.