Comparing corporate financial performance and qualitative information from annual reports using self-organizing maps

Abstract

This paper develops a methodology to extract concepts containing qualitative information from corporate annual reports. The concepts are extracted from the corpus of U.S. corporate annual reports using WordNet ontology and singular value decomposition, and further visualized using self-organizing maps. The methodology makes it possible to easily compare the concepts with future financial performance. The results suggest that annual reports differ in terms of the concepts emphasized reflecting future financial performance.