Abstract: A method for table extraction from journal literature based on text state characteristics is disclosed. The method includes: constructing a table model according to characteristics of tables in journal literature, where the table model includes two parts: a table caption and table content, building a text line set, table detection, table data positioning, table reconstruction, building a cell data set, restoring data of merged cells, checking the cell data set, and outputting table data. The method is particularly designed based on characteristics of tables such as three-line tables widespread in PDF journal literature, which can realize accurate and correct extraction of specific tables in the PDF journal literature, and especially can ensure a logic relationship of a three-line table. The whole process neither requires manual intervention or interaction nor requires table selection, so that the whole extraction process is automatic.

Abstract: A method for table extraction from journal literature based on text state characteristics is disclosed. The method includes: constructing a table model according to characteristics of tables in journal literature, where the table model includes two parts: a table caption and table content, building a text line set, table detection, table data positioning, table reconstruction, building a cell data set, restoring data of merged cells, checking the cell data set, and outputting table data. The method is particularly designed based on characteristics of tables such as three-line tables widespread in PDF journal literature, which can realize accurate and correct extraction of specific tables in the PDF journal literature, and especially can ensure a logic relationship of a three-line table. The whole process neither requires manual intervention or interaction nor requires table selection, so that the whole extraction process is automatic.