In ontology-based applications, the authoring of complex concepts or queries written in a description logic (DL) is a difficult task. An established approach to generate complex expressions from examples provided a user, is the bottom-up approach. This approach employs two inferences: the most specific concept (MSC), which generalizes an ABox individual into a concept and the least common subsumer (LCS), which generalizes a collection of concepts into a single concept. In ontology-based situation recognition the situation to be recognized is formalized by a DL query using temporal operators and that is to answered over a sequence of ABoxes. Now, while the bottom-up approach is well-investigated for the DL EL, there are so far no methods for temporalized DLs. We consider here the temporalized DL that extends the DL EL with the LTL operators next (X) and global (G) and we present an approach that extends the LCS and the MSC to the temporalized setting. We provide computation algorithms for both inferences-even in the presence of rigid symbols-and show their correctness.