Uncertainty handling in document analysis

Automatically performed document analysis gains more and more importance since electronical communication media increase the exchange of information. One important problem in the design of document analysis systems is the development of techniques for the resolution of ambiguities in natural language texts. In order to deal with ambiguity the MULTOS classification system employs a specific technique for reasoning with uncertainty. The paper analyzes different approaches for reasoning with uncertainty with respect to the requirements implied by the document classification task. In particular, it is shown that the knowledge group formalism is adequate for document classification.<>