Czech Insolvency Proceedings: Extraction of Numerical Information and Its Analysis

Abstract

The Czech Insolvency Register contains almost 1500000 of scanned document copies. The paper is focused on efficient extraction of information on the amount of debt from these pdf-files.

Based on the found values of debt and the creditors, individual debtors can be grouped together to form clusters of individuals with a similar structure of debt. Finally, the overall value of debt can be assessed both for the creditors and for the entire country.

Keywords

• data mining
• knowledge extraction
• text processing
• semantics assignment
• classification
• clustering