Fuzzy Semantics in Closed Domain Question Answering

In this chapter we describe fuzzy semantics in a disaster question answering domain. Emergency management implies to go from reports or descriptions of the disaster into rescue work. These descriptions are usually expressed linguistically by witnesses and must be transformed into comprehensible data. The aim of our approach is to achieve a multiple interest: modeling data through (un)balanced fuzzy 2-tuples should permit to decrease the CPU time while keeping an adequate approximation. Another interest lies in the recourse to the natural language processing (NLP) that should offer a multi-lingual interface to non-expert users. Thanks to our method, these users will be helped to express their diagnosis (of the disaster consequences) which will be translated into a dataset representing the catastrophic event.