Bennett Jr, F.G.; (2008) Reconstructing financial statements. Presented at: DESI II: Second International Workshop on Supporting Search and Sensemaking for Electronically Stored Information in Discovery Proceedings, University College London, UK.

Preview

PDF 9137.pdf Download (426kB)

Abstract

This paper introduces a tool for the reconstruction and validation of categorized totals embedded in untrusted and unformatted text, such as OCR scans of nancial statements. The tool is a spino of academic research into the funding of Japanese third-sector organizations, the annual reports of which are frequently published reports in the form of PDF les containing document images. A number of techniques at string- line- and document-level are used to resolve ambiguities and obtain the greatest possible recovery rate for the underlying data, while excluding the content of untrustworthy documents from the nal sample. In a preliminary trial \in the wild", the tool has returned validated income totals for 47.9% of the documents in a heterogeous set of 2205 annual reports.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item