In this paper, we revisit an original proposition on pattern structures for structured sets of attributes. There are several reasons for carrying out this kind of research work. The original proposition does not give many details on the whole framework, and especially on the possible ways of implementing the similarity operation. There exists an alternative definition without any reference to pattern structures, and we would like to make a parallel between two points of view. Moreover we discuss an efficient implementation of the intersection operation in the corresponding pattern structure. Finally, we discovered that pattern structures for structured attribute sets are very well adapted to the classification and the analysis of RDF data. We terminate the paper by an experimental section where it is shown that the provided implementation of pattern structures for structured attribute sets is quite efficient.