Table Content Understanding in SmartFIX

The analysis of table structures and the retrieval of table contents is widely agreed to be a difficult challenge in the area of document analysis systems. Instead of extracting the layout of tables, we are interested in understanding their content. In this paper, we present and discuss the smartFIX approach to table recognition and content extraction. Rather than relying on layout features only, we recognize tables by taking into account the presence and semantics of data entities that we expect to find contained in a table. The relationship of a document, including a table, to a specific business process aids in shaping helpful knowledge and expectations about the table's content. smartFIX is a commercial document analysis system complying with the complete bandwidth of industrial requirements. Therefore, smartFIX must locate the tables and extract its business process relevant information with high reliability.