A Natural Language Processing Approach to Document-based Question Answering

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

DocuMind is a document-based question answering system built entirely on classical NLP techniques, without relying on deep learning.It combines TF-IDF vectorization with cosine similarity to locate relevant passages within uploaded documents (PDF, TXT, DOCX), then applies rule-based logic to extract precise answers based on question type-WHO, WHEN, WHERE, WHY, HOW MANY, WHAT, and YES/NO.Deployed as a Flask web application, the system delivers real-time answers with confidence scores and links back to source context for verification.Evaluation across diverse documents shows that DocuMind achieves solid retrieval precision and fast memory performance, demonstrating that lightweight, feature-engineered NLP approaches remain a practical and efficient alternative to deep learning for document-based question answering.