الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 129 |  |  |
| | | from | 129 | | |
Abstract
The Semantic Web has many features which cannot be handled by using the traditional search engines. Classical retrieval and ranking algorithms are not necessarily suitable for the Semantic Web. It has many features which cannot be handled by using traditional search engines. It extracts metadata for each discovered Web document in RDF or OWL formats, and computes relations between documents. In Semantic Web, the Web documents are ranked according to their contents. These ranking techniques do not depend only on the keywords but also on the contents of the Web documents.
We developed a ranking technique which can be used in a Semantic Web search engine to find a proper ontology for the submitted search query. The proposed ranking technique returns with the most related document from the repository of Semantic Web Documents (SWDs) by using a modified version
of the ObjectRank technique. Then, it creates a sub-graph for the most related SWDs to calculate the hubs and authorities of the processed document by using the HITS algorithm. Our proposed technique increases the quality of the results and decreases the execution time needed for processing the user’s query.
We conducted three different experiments to test the performance of our proposed indexing and ranking technique on the Semantic Web. The first experiment is conducted to test the performance of the proposed ranking algorithm by comparing it with both Swoogle result and human ranking. The resulting values show that the ranking produced by our technique are very close to the ranking produced by the human ranking. The second experiment is run to measure the effect of the total size of the sub-graph on the quality of the result. It show that bigger m, which presents the number of in-links to the sub-graph, implied that more time to construct the sub-graph. Therefore, the quality of our rank algorithm is strongly affected by m. Thus, one has to strike balance between the quality of results and the time needed to construct the sub-graph. Finally, the third experiment is conducted to test the effect of initializing each node with the ratio of all links that the node receives as in-links instead of giving the same initial value for all the pages. We found that this initial hypothesis reduces the consumed time by third.