In today’s global market, applications must support multiple languages, translating an essential yet time-consuming task. Our team experienced firsthand the challenges of building and maintaining translation structures, managing manual translations, and ensuring accurate implementation across applications. To overcome these obstacles, we harnessed our expertise in AI to create a powerful platform that automates and simplifies the translation process.
Discover how our innovative solution not only speeds up multilingual app development but also enhances collaboration with clients, providing a seamless experience from translation to implementation. Dive into our case study to see how we transformed a tedious task into an efficient and streamlined workflow, benefiting both our team and our clients. Authored by Luis Alvarez, Developer at the Cactai Team.
Why are we doing it? What is the pain that we had?
Nowadays almost all apps are intended to work with many languages, we face the painfully time-consuming tasks that usually involve translating an application, for example:
- Building initial translation structure in either a json or xml file;
- Translate every one of the values in that file into all desired languages manually;
- Implementing the translation files in the application following the right format, based on the translation library that the team is using;
- Control de status for the translation of each term;
- Have a reliable version control for updates on the translation files…
So we decided to use our expertise in AI to help with all of these repetitive and manual tasks!
How did we approach the problem? How did we go through the issues that needed to be addressed and cleared them?
- We decided to create a platform where we can manage and tackle all three problems. So firstly we created a Python POC (Proof of Concept) to ensure we could translate a JSON file without changing any property values using AI and… it worked like a charm!
- After solving the main issue (since we only speak English… and Klingon LOL) We faced the remaining issues: building the initial translation structure as a JSON file and applying that translation by changing current texts with the desired translation library calls (in this case we were using Transloco library)
- For this purpose, we created a microservice to handle the scanning of texts on the app to create the initial JSON structure and also to apply newly created properties through the app changing texts for translation library calls.
What did we implement? What tools did we use?
- The decision was easy, split up responsibilities in two different backend APIs.
- One of them is to manage AI integration to translate JSON files or single properties
- Another one is to manage everything related to reading/writing onto files to create the initial JSON translation file and also to change texts with translation library calls. This will also have a connection to a DB to save JSON and change requests from clients.
- Also, we wanted to have a Frontend that was useful, not just for developers to apply translations but also for clients to REVIEW the translations of their projects and be able to change translations from properties if needed. This front end will manage: projects, translation files in projects, and a viewer/editor to see all properties in selected files and change property value requests from clients.
- From Frontend we will also have different modules to trigger needed actions like:
- Section to generate the first translation JSON file and view its result
- Section to change texts in files to translation library call with the right property
- Save translation files, and change requests and projects into a DB so we don’t need to scan every time.
How does it help us in the day-to-day?
This platform is invaluable during the final stages of developing a Minimum Viable Product (MVP) for any app, as it significantly reduces the time spent on translation-related tasks. Additionally, providing clients with a platform to review translations and submit change requests has greatly decreased the response time between us and our clients for these adjustments.
Conclusion
Creating this kind of platforms has helped us to lower a lot the time to market of our apps since it cover tedious tasks and also provides our clients a platform where they can give feedback easily.
