A Three Layer System Architecture for Web-Based Unstructured Data Management

Zhiyong Peng, Hui Wang, Yuwei Peng, Bo Xu, Zeqian Huang

Abstract: With the rapid growing of data on Web, we are facing three serious problems. Firstly, there are a huge number of data resources which are heterogeneous and dynamic.Secondly, most of data on Web are unstructured. Thirdly, there are various kinds of Web users who have different interests and requirements. In this paper, we proposed a new system architecture for unstructured data management on Web to solve these problems by integrating data spaces, database and meta search engine. The system architecture consists of three layers for data gathering on demand, dynamic management and personalized service, respectively. Data servicing layer allows Web users to create data spaces with advanced functions to manipulate and access Web data, eg, cross media query and automatic recommendation. Data managing layer models both Web data and their semantic relationships using our object deputy database named as TOTEM; it also supports schema evolution and dynamic classification. Data gathering layer extracts user's interest from his or her data space; gathers the related data on Web and further analyzes their semantic relationships. Finally, we implemented a prototype system Tmusic based on this new system architecture to show its availability.