This paper presents SHIRI-Querying, an approach for semantic search in semistructured documents. We propose a solution to tackle incompleteness and imprecision of annotations at querying time. This solution relies on two elementary reformulation types that use the notion of aggregation and the documents structure. We present the dynamic algorithm (DREQ) which combines these elementary transformations to construct ordered reformulations of user queries. Experimentations on two real datasets show that reformulations increase greatly the recall and that the precision is better for the first returned answers.