Multilingual Pathology Vusual Question Answering Dataset

The Pathology Visual Question Answering (PathVQA) dataset is a comprehensive collection of 4,998 pathology images paired with 32,799 question-answer pairs. Derived from widely-used pathology textbooks and digital libraries, PathVQA includes both open-ended and binary ("yes/no") questions related to pathology images. Initially in English, the dataset was curated and obtained from the Hugging Face dataset library. To enhance multilingual adaptability, we translated the dataset into French, Hindi, and Yoruba using Neural Machine Translation (NMT) and backtranslation techniques, ensuring consistency across the additional language versions. This dataset offers a valuable resource for advancing machine learning models in medical and pathology-related visual question answering tasks.