Case representation is the basis for case-based reasoning. The representation should efficiently organize large amounts of information which is usually unstructured information, related to the emergency cases. This paper describes the attributes that should be included in emergency case descriptions with a case framework based on analyses of emergency events and information with both structured and unstructured information. The structured information uses fuzzy sets to describe the non-quantitative data using memberships instead of numerical values. The unstructured information is analyzed by three algorithms for keyword extraction with the word co-occurrence approach chosen as the best. The extracted keywords are then used to obtain the proper information segments for the unstructured part of the emergency case. The whole case is represented in a semi-structured form with quantitative attributes used in the case retrieval and text data used in the case reasoning. The results show that this approach gives good results as a foundation for case-based reasoning applications.