To speed up queries over XML and semi-structured data, a number of structural indexes have been proposed. The structural index is usually a labeled directed graph defined by partitioning nodes in the XML datagraph in to equivalence classes and storing equivalence classes as index nodes. On the basis of the Inter-Relevant Successive Trees (IRST), we propose an efficient adaptive structural index, IRST (k)-index. Compared with the previous indexes, such as the A (k)-index, D (k)-index, and M (k)-index, our experiment results show that the IRST(k)-index performs more efficiently in terms of space consumption and query performance, while using significantly less construction time.