Bilingual Handwritten Indian Language Translation

Resumo

This paper presents a deep learning-based system for translating handwritten Sanskrit text into English. The system addresses key challenges posed by Sanskrit, including complex grammar, flexible sentence structure, and varied handwriting styles. The pipeline begins with image preprocessing to enhance handwritten text clarity, followed by Optical Character Recognition (OCR) to convert the images into machine-readable Sanskrit text. A Sequence-to-Sequence (Seq2Seq) model using Long Short-Term Memory (LSTM) networks,

enhanced with an attention mechanism, then translates the text into English. The attention mechanism enables the model to focus on relevant parts of the input during translation. Translation quality is evaluated using standard metrics such as ROUGE, Precision, Recall, and F1 score. Experimental results demonstrate the model’s effectiveness in producing accurate translations. This work contributes to machine translation for low-resource languages and supports the preservation and accessibility of ancient cultural texts.