Abstract. In database applications, the availability of a conceptual schema and semantics constitute invaluable leverage for improving the effectiveness, and sometimes the efficiency, of many tasks including query processing, keyword search and schema/data integration. The ObjectRelationship-Attribute model for Semi-Structured data (ORA-SS) model is a conceptual model intended to capture the semantics of object classes, object identifiers, relationship types, etc., underlying XML schemas and data. We refer to the set of these semantic concepts as the ORAsemantics. In this work, we present a novel approach to automatically discover the ORA-semantics from data-centric XML. We also empirically and comparatively evaluate the effectiveness of the approach.