Elevating Partner Content Through An Iterative LLM-driven Approach
Track: Data Solutions | DS1 | Intermediate |
Wednesday, June 4, 2025, 9:00am – 9:30am
Held in: Live 1
Presenters:
Veronica Carioni - Skyscanner
Marta Castello - Creative Words
Host: Donna Parrish
In today’s attention-driven economy, where businesses are primarily competing for people’s focus, large language models (LLMs) are emerging as solutions to illuminate previously overlooked or unsupervised content areas.
This presentation details a collaborative effort between Creative Words and Skyscanner to leverage an AI-powered workflow that assesses the quality and linguistic accuracy of high-volume partner-generated content. The project, which is part explores the integration of a multi-layer configuration with an agnostic validation structure to improve MT raw output quality while identifying key pain points in the analyzed texts. Instead of striving for an impractical 100% precision, this framework prioritizes a scalable, unbiased, and iterative quality assurance process to enhance accuracy and ensure replicability in MT results.
This research is presented as part of a broader set of initiatives that reflect Skyscanner’s AI maturity journey and Creative Words’ comprehensive portfolio in linguistic technology integration. Ongoing at the time of this abstract submission, the project will be presented as a step-by-step process, highlighting findings from a contrastive analysis of pre- and post-LLM review quality, as well as performance variations across languages.
Key Takeaways:
- How and why 100% success with LLMs is a no-go
- Real case study involving results of LLMs integration and validation structures
- The broader impact of experimentation
