Header metadata extraction from semi-structured documents using template matching

With the recent proliferation of documents, automatic metadata extraction from document becomes an important task. In this paper, we propose a novel template matching based method for header metadata extraction form semi-structured documents stored in PDF. In our approach, templates are defined, and the document is considered as strings with format. Templates are used to guide finite state automaton (FSA) to extract header metadata of papers. ne testing results indicate that our approach can effectively extract metadata, without any training cost and available to some special situation. This approach can effectively assist the automatic index creation in lots of fields such as digital libraries, information retrieval, and data mining.