In the domain of Computing with words (CW), fuzzy linguistic approaches are known to be relevant in many decision-making problems. Indeed, they allow us to model the human reasoning in replacing words, assessments, preferences, choices, wishes, etc. by ad hoc variables, such as fuzzy sets or more sophisticated variables. In this thesis, we present a new fuzzy representation model to deal with unbalanced linguistic term sets that allow us to handle data with precision and accuracy. This model is based on our fuzzy semantic 2-tuples that we introduce. We apply these semantic 2-tuples to perform a fuzzy semantic interpretation of words in a natural langage dialog context for a geolocation application. In this thesis, we start from a concrete geolocation problem : how to configure the devices that track the mobiles and how to set up the alerts related to the tracking ? The idea is to offer the possibility to go from the end-user business-level objectives expressed through linguistic parameters (via a natural language dialogue) to an appropriate combination of technical parameters. We show how to improve a user interface to offer a qualitative processing instead of a quantitative one. We define theoretical extensions in the CW framework : a model, based on semantic 2-tuples that are introduced, permit to represent linguistic term sets with accuracy and precision, even if they are very unbalanced. These semantic 2-tuples can interpret semantically the user's linguistic terms during the dialogue, and attach a contextual fuzzy semantics to them.
