Bilingual Bangla OCR for Rural Empowerment: Detecting Handwritten Queries and Agricultural Assistance

Abstract

Farmers in rural areas often struggle to access crucial agricultural information due to language barriers, low literacy rates, and limited exposure to digital tools. While many can write in Bangla, most agricultural resources are available only in English or require navigating complex systems, making it difficult for them to find relevant information. Existing Optical Character Recognition (OCR) technologies, which could help bridge this gap, are primarily designed for printed text and often fail to recognize handwritten Bangla script accurately. Issues such as biased datasets, diverse handwriting styles, and background noise further reduce accuracy, making these systems unreliable for real-world use. To tackle these challenges, we have developed a lightweight and unbiased OCR model specifically for handwritten Bangla text. Our solution integrates a custom Convolutional Neural Network (CNN) with InceptionV3, enhancing recognition accuracy while ensuring efficiency for low-resource devices like smartphones. Additionally, we incorporate a two-way translation feature, enabling seamless Bangla-to-English and English-to-Bangla conversion. This allows farmers to write in Bangla, translate content when needed, and access critical information in a way that best suits them. Our solution empowers rural farmers by enabling them to interact with digital platforms in their native language, bridging the gap between handwritten communication and modern technology. Beyond agriculture, this technology has far-reaching applications in tourism, healthcare, education, and government services, fostering digital inclusion. By advancing OCR for Bangla, our research promotes equitable access to technology, equipping communities with essential tools to improve productivity and quality of life in the digital era.