Research Of Keyword Search Model Over RDF Data Graph

As one of the most effective techniques of obtaining useful information from massive RDF data, RDF data search is playing an important role in the domain of data management. However, there still exist some drawbacks in the study of RDF data search. Traditional methods always focus on keyword searching, and ignore the semantic information of the keyword and the structural information among data, resulting in redundant and deviated results. Under the background of the application of extensive data, how to address the problem of outputting query answer sequentially is still urgent. In order to work out the practical problems mentioned above, this thesis concentrates on some points below.Firstly, a novel keyword search retrieve model, which is based on RDF data graph, is proposed in order to improve the existing keyword search-based retrieve techniques and systems, This model pre-processes the RDF data first, and takes it as the basis of result ranking. After searching the graph, the result sub-graphs are generated. At the result ranking stage, these result sub-graphs are scored comprehensively and output in descending order by using top-k method.Secondly, in the research of graph search algorithm, the iterative network-motif detection algorithm is introduced into the graph search algorithm, and a novel iterative RDF data graph retrieve algorithm is proposed. Compared with the traditonal graph search algorithm, the executive efficiency of this method has been greatly enhanced.Thirdly, in order to address the problem of redundant and deviated results existing in RDF data graph, a new result sub-graph similarity computing method (SimLM) is presented. Meanwhile, considering the semantic and stuctual information in RDF data, the stuctual similarity between keyword graph and result graph, and the language similarity are both combined into a new similarity computing method. Using this novel method, much more desired results are preseted eventually.Finally, some experiments are carried out in order to test the validity of this graph similarity computing method (SimLM), The result of the contrast experiment shows that the retrieve model and ranking method perform much better than the traditonal model in respect of consistency and relevance.