In the Semantic Web, the types of resources on the Web and the semantic relationships between resources are defined in an ontology. By using those information, the accuracy and the efficiency of the Web search can be improved. In this dissertation, we focus on the effective semantic search using the ontology.
The goal of the semantic search is to retrieve $\It{top-k}$ resources which are the most relevant to query keywords through semantic relationships in the ontology. To do this, we first propose an effective ranking method for the semantic search. We first devise a measure to determine the weight of the semantic relationship. Based on this measure, we devise a novel relevance scoring function to be used for ranking resources. The novel relevance scoring function considers the number of meaningful semantic relationship paths connecting the resource with keywords, the coverage of keywords related to the resource, and the distinguishable power of keywords related to the resource. Moreover, in order to improve the efficiency of the search, we prune the unnecessary search space using thresholds based on the length and the weight of the semantic relationship path.
For a huge number of resources, it is impractical to compute the relevance scores for all the resources and extract the most relevant $\It{k}$ resources which have the highest relevance scores in a reasonable time. Therefore, we propose an efficient $\It{top-k}$ query processing technique based on a more accurate score estimation. In order to efficiently and accurately estimate the upper bound score of a resource, we devise a bloom filter embedded-index structure called the index filter built on a partitioned input list. Then, we adapt TA-style $\It{top-k}$ query processing algorithms by employing the index filter, so that non-answer candidates are pruned as rapidly as possible. Consequently, the performance of $\It{top-k}$ query processing is considerably improved while preserving the correctness of...