Structures heterogeneity in Web resources is a constant concern in element retrieval (i.e. tag retrieval in semi-structured documents). In this paper we present the SHIRI 1 querying approach which allows to reach more or less structured document parts without an a priori knowledge on their structuring.