Design And Implementation Of A Visual Mining System On Linked Data

With the development of Semantic Web, Linked Data are getting increasingly popular. However, with the volume of Linked Data getting bigger and bigger, the access to the knowledge from Linked Data are becoming increasingly difficult, especially when the user lack the prior knowledge of the Semantic Web. Therefore, it will be useful to provide users with a visualized, customized mining services. In Linked Data's mining task, detection of community is one of the important research areas that can help user understand time-related changes in Linked Data. But the research of community evolution faces two research questions:First, Linked Data contains not only a wealth of link information, but also contains a certain amount of linguistic information in the form of short texts. It is important to consider how to use linguistic information of objects in the research of community evolution. Secondly, how can the detection of community evolution be customizable, visualized and interacted with users, is another problem to be solved.In this thesis, a visual mining system named LIV (LInked data Visualization) is designed and implemented. Meanwhile, the contributions of this paper lie in the following aspects: First, based on the notion of Virtual Document, a text extension operation will be performed on the local linguistic information of each object according to neighboring links from and to the object.Subsequently, segmentation and clustering operation will be performed on Linked Data to generate a series of clustering tree and summarization of each cluster. Finally, we design and implement a scalable link data visualization mining system based on Browser/Server architecture. The system can help users explore the Linked Data easily.LIV is sound in scalability. Users can continue to add new mining module and visualization components to it according to the characteristics of various Linked Data and the requirement of users. Therefore, this paper works on visual data mining on Linked Data, especially proposes a new research ideas and methods on visualization of community evolution detection.