Situation-Aware User's Interest Prediction aims at enhancing the information retrieval (IR) capabilities by expanding explicit user requests with implicit user interests, to better meet individual user needs. However, not all user interests are the same in all situations, especially for the case of a mobile environment. Thus, user interests are complex, dynamic, changing, and even contradictory. Consequently, they should be adapted to the user's specific search context. In this paper, we introduce a new approach that aims at building a dynamic representation of the semantic situation of ongoing mobile environment retrieval tasks. The semantic situation is then used to activate different classification rules of user's past interests at run time. Doing so, the best interest class's is proposed to expand the user's request. Our approach makes use of a semantic enrichment using Dbpedia, providing enriched descriptions of the semantic situations involved for discovering user interests, and enabling the definition of effective means to related contexts. Carried out experiments, undertaken versus Google search, emphasize the relevance of our proposal and open many promising issues.