A novel architecture for knowledge mining from digitised document libraries

This paper examines a novel knowledge mining architecture based on the Azure cloud data and AI services, to extract data from the Emporium library, a modern art journal published between 1985 and 1964. The knowledge mining starts with Optical Character Recognition (OCR) and custom Name Entity Recognition (NER) on digitised images of the pages and provide the final user with an user-friendly search portal to navigates the hundreds of pages in milliseconds through a semantic query. The study proved how this architecture fits from an art scholar’s perspective and how it enables to build more comprehensive statistics and description of the document corpus.