Automatically Identifying Records from the Extracted Data Fields of Genealogical Microfilm

Imagine the ability to search thousands of microfilm documents at once. Yet extracting and organizing this information by hand is nearly impossible. This research proposes an algorithmic process to identify and extract records automatically from tabular tables found in images. The proposed algorithm extracts record patterns by using knowledge of the structure and geometry of tables and a genealogical ontology. The algorithm accepts raw data collected from zoned microfilm images. It receives this data as an XML input file that describes the coordinates of each table cell, the printed text in each cell, if any, and whether or not each table cell is empty. The process produces record patterns that express the geometry and attributes of the records within a table.