TechPrufer
Get started
Grab
Grab logoGrab · Feature trail

In-house Translation Service

1 post · May 2019 – Aug 2024

What this trail covers

Grab has evolved its translation capabilities by developing an in-house translation service to improve quality and cost efficiency. This involved implementing on-the-fly language detection using Lingua, with a fallback to a third-party service. To enhance translation quality, a custom lightweight translation model was trained using synthetic data generated from a larger LLM, incorporating domain-specific context and a post-translation quality check to mitigate issues like number mutation. Caching was implemented to reduce costs, and system messages were optimized for translation.

All Grab trails
  1. How we improved translation experience with cost efficiencyThis post details the evolution of Grab's translation service by implementing on-the-fly language detection using Lingua, with a fallback to a third-party service. It describes the process of training a custom lightweight translation model using synthetic data from a larger LLM, incorporating domain-specific context and a post-translation quality check to mitigate number mutation issues. Caching strategies and optimizations for system messages were also introduced to reduce costs.
  2. Tourists on GrabChat!This post details the use of Grab's in-house translation solution within GrabChat, specifically highlighting its utility for tourists. It showcases how the translation feature, along with chat templates and image sharing, facilitates communication. The post provides data on translation usage patterns by tourists across different countries and languages, and mentions the use of Lingua for language detection and a custom lightweight translation model trained with synthetic data.