| With the rapid increasing of data assets in our society, the information overloadproblem has been paid much more attentions. Especially, accessing semantically correctdata from the massive data assets to support decision-making has become a majortechnology bottleneck. Since relational database is widely used in every walk of life,people focus on accessing data with semantic correctness from distributed andautonomic relational databases.In order to solve problems above, we put forward a new semantic correctnessoriented integrated data access method in this dissertation. We conduct in-depth studyabout the compromise between expressive capability and reasoning complexity ofontology, semi-automatic construction of semantic mapping, semantic query processingand instance checking, and maintenance of integrated data access in dynamicenvironment. We are aiming at providing an innovative idea of converting informationadvantage into decision superiority, and the contributions of this dissertation can besummarized as below.Firstly, considering the inherent contradictory between the expressive capabilityand reasoning complexity of ontology, we proposed and demonstrated a layer-dividedTBox model and a dynamic ABox model. These models will help us depress thereasoning complexity according to a divide-and-conquer strategy. After that, weextended a description logic namedDL-LiteONWLon the basis of DL-Lite family. Thisnew description logic can describe the semantics of ER model sufficiently, and also hasproper data complexity. Basing on studies above, we establish a novel integrated dataaccess method, and present the basic architecture and procedure of our method.Secondly, in order to improve automation of mapping construction and reduce thecomplexity of query answering, we adopt a new mapping formalism namedLAV+O-GAV to describe the relations between ontology and relational database.During the process of constructing LAV mappings, we offer a novel mapping discoverymethod based on concept connected graph forDL-LiteNOWLontology. This method candiscover semantics hidden in relational schema in a semi-automatic way. Theconstruction of O-GAV mappings based on layer-divided TBox and LAV mappings,and the function of O-GAV mappings is building a bridge between the intensionalknowledge segment and extensional fact stored in relational database. Our experimentshowed that adopting LAV+O-GAV manner will speed the development of mappingconstruction on the premise of ensuring the semantic correctness of mappings.Thirdly, we provide a virtual object oriented query processing method based onlayer-divided TBox model, dynamic ABox model, and O-GAV mappings. This method consists of three phases: First, according to characters ofDL-LiteNOWLontology andrewriting rules, we used SuperRef method to rewrite user query into a set of conjunctivequeries; Second, dealing with queries obtained from rewriting step based on O-GAVmapping, and a dynamic ABox will formed after this phase; After that, checkinginstances according to all axioms in current layer of TBox, and present the objectswhich satisfied all constraints. Experiment shows that our query processing method hasbetter stability, and it can meet the efficiency demands of data access.At last, this paper presents an ontology evolution method called LTOOE for layerdTBox model. This method can keep the layers of TBox unchanged after ontologyevolution, and limit the influence that may be imposed on the first layer of TBox.Further more, considering the effect of ontology change transmission, we providemethods of detecting invalid LAV mappings and illegal LAV mappings, and proposedan incremental maintenance method for invalide LAV mappings.To sum up, our SCIDA method can provide semantically correct data for users andintelligent applications, and it makes sense to research field such as data integration,data access, and intelligent decision-making and semantic web. Meanwhile, we providea feasible way to exploit and utilize data semantics implicated in the relational database. |
PDF Full Text Request