Attributes difficult to describe are managed by using a long character string attribute dimensionality in a unified manner generally when databases of existing management systems are designed. As data application needs deepen, the mode has some problems in data similarity retrieval. Starting from the problem point, the problem is elevated and described as a similarity matching problem of universal order-independent item groups, and the invention provides an effective solution for quick similar semantic matching of the order-independent item groups. Main ideas of the method include: designing an effective numbering mode to convert order-independent item group codes originally composed of attribute names into order-dependent item groups composed of serial number values; performing quick matching calculation on the order-dependent item groups.