Tables, a primary modality for organizing and presenting information for human comprehension, are ubiquitously found in documents. Their design poses significant challenges for systems, including large language models, when it comes to processing and understanding tabular data. We propose a novel method to free the tabular data encumbered inside documents (PDFs, HTML pages, Word documents, etc.) and perform question-answer (QA) on this data via natural language interaction. Our method stresses on its domain-agnostic and “open”-QA-oriented abilities, that exceed the performance of current LLMs in certain situations. We achieve this using a combination of table extraction tools, followed by the creation of a knowledge graph using the tabular data sources and employing QAnswer ( https://www.qanswer.eu/ ), a QA system generator. A video demonstration showcases our tool’s capabilities on United Nations (UN) disability documents and webpages.