Dementia Education Chatbot Sources and Study Results

This deposit contains the core reproducibility materials for the **AI4HOPE Dementia Companion** project. It includes the merged multilingual source metadata used to build the retrieval index, the frozen retrieval index itself, and the anonymized study result tables from the expert and patient/caregiver evaluations. What is included - Merged source metadata: a unified parquet file created by combining and preprocessing crawl outputs from multiple language-specific runs.- Frozen retrieval index: the exact index state used in the study to support reproducibility of the retrieval-augmented generation workflow.- Study results: anonymized CSV/Excel tables containing outcomes from the expert study and the patient/caregiver study. Why this deposit exists The goal of this deposit is to support transparency, reproducibility, and reuse. The merged metadata retains provenance fields such as source language, source file, source folder, and source variant, making it possible to inspect how the corpus was assembled and how the retrieval index was built. Use and interpretation This deposit is intended for research and reproducibility purposes only. External source material remains subject to the licenses and terms of the original sources. Any participant-related study data should be used in accordance with the ethical and privacy constraints under which the study was conducted. Associated project AI4HOPE Dementia Companion Chatbot is a multilingual retrieval-augmented generation system designed to support dementia care information for related but distinct study settings. Code Repo https://github.com/HUMADEX/ai4hope-djc-app HuggingFace data https://huggingface.co/datasets/HUMADEX/dementia_education_chatbot_sources