Generative AI-based translation tools suite

Thumbnail Image

Date

2026-01

Journal Title

Journal ISSN

Volume Title

Publisher

BRAC University

Abstract

This project presents the design and development of an intelligent document translation and preview system capable of handling multi-format files including Word, Excel, text, and images. The system integrates Optical Character Recognition (OCR), automated translation, and interactive preview functionalities within a unified interface. A dynamic tab-based architecture allows users to manage multiple translation sessions simultaneously, with real-time state management ensuring synchronization between the client and server. The system leverages modern large language models (LLMs), such as the Gemini API, to perform context-aware translations. To enhance efficiency, the backend implements asynchronous batching and batched API calls, minimizing latency during large-scale translation tasks. The chat assistant module provides contextual editing support and instruction-based translation refinement, allowing users to iteratively improve outputs without manual re-uploading. Advanced preview mechanisms have been implemented for each file type: formatted text display for Word, table parsing for Excel, OCR-based text extraction and alignment for images, and editable text containers for plain files. Together, these modules form a cohesive, scalable, and extensible framework for intelligent multilingual document processing. The result is a technically robust system that demonstrates a practical balance between automation, user control, and performance optimization

Description

Cataloged from PDF version of thesis.
Includes bibliographical references (page 22).
This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2026.

Keywords

Document translation, Large Language Models (LLMs), Gemini API, Optical Character Recognition (OCR), Web application

Citation

Endorsement

Review

Supplemented By

Referenced By