Visit Lionbridge Games

SELECT LANGUAGE:

Optimized translations with LLM-assisted post-editing.

Lionbridge LLM-Assisted Post-Editing: The New Standard

How Lionbridge’s LLM-assisted post-editing solution can save you up to 40% on your overall translation spend

As Artificial Intelligence (AI) technology rapidly advances, enterprises like yours pursue innovative generative AI (GenAI) language solutions to cut costs, maximize existing budgets, and gain a competitive edge. Separating the hype and empty promises from proven solutions is crucial with an explosion of available options.

  • Which solutions consistently deliver on their promises?
  • Who is transparent about AI technology’s current usefulness and limitations?
  • Who has the experience and a proven track record of extensively testing and validating their offerings?

Lionbridge’s Large Language Model (LLM)-assisted post-editing offering is part of a comprehensive translation solution delivered via the industry’s only global content management platform built with an AI-first approach. Much more than a Translation Management System (TMS), the Lionbridge Aurora AI™ platform provides automated decisioning and flexible workflows to minimize human interventions end-to-end.

Lionbridge’s LLM-assisted post-editing solution—orchestrated by Aurora AI—reduces the content requiring manual review and is essential for achieving optimal translation efficiency and cost savings.

What Is LLM-Assisted Post-Editing?

LLM-assisted post-editing uses GenAI/LLMs to refine and improve the output of Machine Translation (MT) engines.

As a first step, content is translated using Translation Memories (TMs), if available, and Neural Machine Translation (NMT) engines. Subsequently, the LLM undertakes the first pass at the post-editing step, which humans have traditionally performed. Due to the generative AI models’ ability to grasp context, nuance, and linguistic subtleties, LLM-assisted post-editing reduces the need for extensive manual reviews.

Why Not Use LLMs for Translation Instead of Post-Editing?

Currently, LLMs cannot match the translation speed and affordability of top NMT engines, as Machine Translation (MT) has evolved over the years. This reality firmly secures Machine Translation a spot in the translation workflow—at least for now.

Using MT for initial translations at the beginning of the process and LLMs for post-editing and quality assurance tasks later in the translation workflow leverages the best of both technologies for improved translation efficiency and very good translation results.

Person using a laptop.

How Does Lionbridge’s LLM-Assisted Post-Editing Solution Work?

The translation workflow starts by leveraging existing Translation Memories and generating a machine-generated translation of the new content. The LLM then evaluates the translated content. During the Assisted Post-Editing steps, the LLM analyzes the text, identifies areas for improvement, and makes the necessary adjustments. Lastly, the LLM identifies text that requires further review by a human.

Human post-editors focus on fine-tuning and quality assurance rather than extensive rewriting. A more thorough manual post-edit review by a human in the loop is always an option for content that warrants an extra polish. (Learn more about human-in-the-loop AI translations.)

What Are the Benefits of Lionbridge’s LLM-Assisted Post-Editing Solution?

Lionbridge’s LLM-assisted post-editing solution significantly reduces the content that requires manual review while reviewing and improving Machine Translation output.

By decreasing human effort, companies get faster turnaround times and can lower the costs of translating new content by up to 40%. Companies can localize larger volumes of content more efficiently, allowing them to scale operations without compromising quality.

How Effective Is Lionbridge’s LLM-Assisted Post-Editing Solution?

Lionbridge is using its LLM-assisted post-editing solution on hundreds of language pairs. Manual evaluations of random project samples conducted by humans have shown that the solution produces very good results across a diverse range of language pairs.

Our continuous monitoring and performance assessments ensure that our LLM-assisted post-editing solution delivers the desired results. We guide our customers to use the languages best suited for LLM-assisted post-editing as part of our ongoing commitment to sell only solutions that work.

Lionbridge LLM-Assisted Post-Editing at a Glance

Having deployed our LLM-assisted post-editing solutions for a substantial number of customers over the past year, we’ve gained ample experience working with the technology so you can feel confident about using it for yourself.

Key facts:

  • Live since November 2023
  • Continuous monitoring and evaluation
  • Used by over 1,000 Lionbridge customers with post-editing projects to date and growing
  • Support for 378 language pairs
  • Quick deployment (within days without the need for LLM training)

What Sets Lionbridge’s LLM-Assisted Post-Editing Solution Apart?

Lionbridge’s LLM-assisted post-editing solution delivers unparalleled efficiency and accurate results. Here’s why it’s a standout solution:

  • Advanced AI Models: We leverage state-of-the-art, leading-edge LLMs that do not require additional training. In addition, our open architecture allows us to use different LLMs depending on customer requirements.

  • Comprehensive Language Support: The solution supports 378 language pairs, making it one of the most versatile options on the market. You can turn to us if you need translations for widely spoken languages like English, Spanish, and Mandarin or less common languages.

  • Customization and flexibility: We offer tailored prompting, enabling us to adjust the AI model’s behavior to fit specific customer requirements. This flexibility ensures that the translations adhere to the desired tone, style, and context, making them suitable for various types of content.

  • Integration with Lionbridge Aurora AI: Our LLM-assisted post-editing solution is a key component of Aurora AI. This AI-first platform provides an automated end-to-end solution for LLM-driven content creation and localization.

"You can’t embrace the new AI world with a platform paradigm dating back 20 years ago. Instead of using AI as a buzzy band-aid for legacy platforms, we’ve built a platform that puts AI first and enables humans to curate the process, including LLM-assisted post-editing."

–Vincent Henderson, Lionbridge VP of Strategy

person reviewing complex data sets

When Should Enterprises Use LLM-Assisted Post-Editing?

The time to use LLM-assisted post-editing is now. Use the solution for all or specific content to reduce your translation spend and get savings. With over a year of experience with this solution, we know what works and what doesn’t. We can help you execute a staggered deployment and ensure we deploy the best-suited solution to each content type.

Why Lionbridge?

When working with Lionbridge, you can expect us to dig deep into your needs. Instead of simply responding to a language request, we take the time to explore the goals underlying your request. This extra upfront step ensures we propose the best translation technology solution tailored to your needs.

Moreover, we only promise things we can confidently deliver. We build our services and solutions on a foundation of TRUST. Our TRUST framework is comprised of the following elements:

  • T, for transparent
  • R, for reliable
  • U, for Useful
  • S, Scalable
  • T, Timely

Rest assured, when we say we will deliver on something, you can count on it thanks to our human expertise, which enables us to control the automation we put forth, our decades of AI and language delivery experience, and the remarkable results we are achieving using cutting-edge AI.

Get in touch

Discover how easy it is to save up to 40% on your overall translation spend. Contact us now to get started.

  • #technology
  • #translation_localization
  • #ai
  • #generative-ai
  • #blog_posts
linkedin sharing button

AUTHORED BY
Janette Mandell

Let's Talk

Business Email Only