Segmentation of Document Using Discriminative Contextfree Grammar Inference and Alignment Similarities

Abstract

Abstract-Text Documents present a great challenge to the field of document recognition. Automatic segmentation and layout analysis of documents is used for interpretation and machine translation of documents. Document such as research papers, address book, news etc. is available in the form of un-structured format. Extracting relevant Knowledge from this document has been recognized as promising task. Extracting interesting rules form it is complex and tedious process. Conditional random fields (CRFs) utilizing contextual information, hand-coded wrappers to label the text (such as Name, Phone number and Address etc). In this paper we propose a novel approach to infer grammar rules using alignment similarity and discriminative context-free grammar. It helps in extracting desired information from the document.