.svg){kind=logo}
AI Solutions
Additional Services
Case Study: Multilingual Retail Marketing
New AI Content Creation Solutions for a Sports and Apparel Giant
- RESOURCES
“The targeted application of style guides and language instructions into LLMs is crucial to boosting translation quality from 80% to 99%. We collaborated with our customer to implement dynamic prompting based on context and content type in their own LLM system for hyper-personalization. This capability is also readily available in the Lionbridge Aurora AI™ platform, enabling any company to speak more intimately to their target audiences through various techniques, such as offering 20 different versions of Spanish or tailoring their tone for more demographic segments.”
—Vincent Henderson, Lionbridge Vice President of Strategy
Accurate and consistent translations are essential for global communication. Large Language Models (LLMs) play a pivotal role in this undertaking with their ability to enhance the speed and cost-effectiveness of translation processes. However, model performance significantly depends on how well the model is trained to apply a brand’s voice and style. Training LLMs, which involves prompt engineering, is a complex but essential task for companies looking to optimize their automated translations.
A leading hyperscale technology company engaged Lionbridge for AI data services to improve the quality of its LLM’s output for website localization and marketing email template translations. To achieve their objective, the customer needed its in-house LLM to be trained to follow a detailed 100-page style guide (encompassing more than 500 rules) across 47 languages within an aggressive 6-month timeline.
Lionbridge extracted language rules from the customer’s extensive style guide, turning them into LLM prompts and enforcing their applicability in AI translations to achieve improved generative AI output.
At the outset of the engagement, Lionbridge conducted a thorough review and analysis of the customer’s existing style guide and collaborated with Subject Matter Experts (SMEs), linguists, engineers, and project managers to ensure effective LLM training.
Read the case study to learn:
- What our three-phase solution entailed.
- The essential role of a vast community in successfully training an LLM.
- How to achieve hyper-personalization without an in-house LLM. (Most companies don’t have one.)
This case study demonstrates the crucial importance of integrating a carefully curated and streamlined set of style guide rules into an LLM through prompt engineering for enhanced translation quality. This advancement means that traditional style guides will be phased out.
Lionbridge’s AI data services, which train an LLM to apply style guide rules, are suitable for enterprises with an extensive, well-defined style guide, whether or not they operate an in-house LLM. Lionbridge can train a company’s LLM, as we did for this customer. Alternatively, we can train our LLM to ingest and apply a customer’s complex style guide via the Lionbridge Aurora AI platform and our advanced Assisted Post-Editing (APE) offering.
Fill out our contact form to start a conversation with us.
We’re eager to understand your needs and share how our innovative capabilities can empower you to break barriers and expand your global reach. Ready to explore the possibilities? We can’t wait to help.
