About Project Udaan
Problem statement of this project
Research & Insights
To tackle the problem, I started with thorough research – both secondary research (desk study) and primary research (user study) – to understand the landscape of translation and the users' needs.
Landscape Study: I analysed existing translation tools and workflows. Academic research pointed out that many current Computer-Aided Translation (CAT) tools rely heavily on translation memory and glossaries, but still require extensive post-editing by humans.
User Shadowing & Interviews: I then spoke with and observed 20+ professional post-editors who were involved with the textbook translation-editing initiative using the Udaan Tool.
Here are the key insights from my research
To tackle the problem, I started with thorough research – both secondary research (desk study) and primary research (user study) – to understand the landscape of translation and the users' needs.
Note: most post-editors using this tool were over 40 years old, meaning they were dealing lower vision and digital literacy (though well-versed with computer applications).
Here are the problems are discovered from the interview sessions with the post-editors from All India Council of Technical Education (AICTE)
Concept Proposal and feature implementation
Categorization and accessibility
Innovative features streamline and accelerate the post-editing workflow
This section acts as a centralized repository for all non-text elements in the textbook, including tables, figures, and equations. Editors can add, save, import, and filter these elements across pages, with search support for quick retrieval.
By decoupling elements from individual pages, the system allows editors to reuse the same table, figure, or equation wherever it appears in the book. This reduces repetitive work, prevents inconsistencies, and makes it significantly easier to manage complex technical content at scale—especially in long engineering textbooks where the same elements are referenced multiple times.
Text entry was simplified to reduce friction while translating in Indian languages. Editors can use speech-to-text to dictate phrases directly into the document instead of typing, which is especially helpful when working with complex scripts or long sentences.
The feature supports multiple Indian languages and gives users full control over the audio input—they can re-record, delete, replay, or confirm the transcription before inserting it. This allows post-editors to focus on accuracy and flow rather than struggling with keyboards or input methods, making the translation process faster and more accessible.
