Intelligent Dictionary Elements Extraction using Optical Character Recognition and Deep Learning based Object Detection

The Sanskrit Encyclopaedic Dictionary conceived and compiled by Deccan College Post Graduate Institute, Pune, India is available in paper format. For greater reach, there is a need for digitization of the dictionaries and conversion of these volumes into a comprehensive structured database. These volumes have Articles with a typical complex structure. In this paper, a novel approach for Automatic Lexical Element Extraction from Sanskrit Encyclopaedic Dictionary Pages (ALEE-SED) that consists of Sanskrit Dictionary Page Segmentation into articles and their meaning blocks using Optical Character Recognition (OCR) and Detectron2 Model further followed by an intelligent article wise dictionary (Lexical) element extraction is described.