Document Layout Analysis for Semantic Information Extraction

摘要

Using machines to automatically extract relevant information from unstructured and semi-structured sources has practical significance in todays life and business. In this context, although understanding the meaning of words is important, the process of identifying self-consistent geometric and logical regions of interest-blocks, cells, columns and tables, as well as paragraphs, titles and captions, only to mention a few-is of paramount importance too. This complex process goes under the name of document layout analysis. In this work, we discuss newly designed techniques to solve this problem effectively, by combining both syntactic and semantic document aspects. These techniques described here are at the basis of KnowRex, a comprehensive system for ontology-driven Information Extraction.