Abstract
Semantic analysis extracts semantic information from natural language texts and endeavors to make implicit facts explicit. Context and experience - in terms of previously achieved knowledge - are essential to solve this task. Confident semantic information from ambiguous natural language can only be obtained if set in a sufficient context. Conventional Named Entity Mapping algorithms use context as positive example environment for the disambiguation process. Traditional machine learning algorithms also apply negative examples to train a classifier for a specific subject. For Named Entity Mapping this can trivially be achieved by manual curation of black lists. These black lists contain entities that do not make sense in the given context. This paper describes an approach how to achieve a negative context dynamically during the disambiguation process and how to make use of this negative context for subsequent analysis steps.