| The need for efficient solutions to the problem of disseminating and sharing of data is growing. Digital libraries provide an efficient solution for disseminating and sharing large volumes of data to diverse sets of users. They enable the use of structured and well defined metadata to provide quality search services. Most of the digital libraries built so far follow a centralized model. The centralized model is an efficient model; however, it has some inherent problems. It is not suitable when content contribution is highly distributed over a very large number of participants. It also requires an organizational support to provide resources (hardware, software, and network bandwidth) and to manage processes for collecting, ingesting, curating, and maintaining the content.; In this research, we develop an alternative digital library framework based on peer-to-peer. The framework utilizes resources contributed by participating nodes to achieve self-sustainability. A second key contribution of this research is a significant enhancement of search performance by utilizing the novel concept of community evolution. As demonstrated in this thesis, bringing users sharing similar interest together in a community significantly enhances the search performance. Evolving users into communities is based on a simple analysis of user access patterns in a completely distributed manner. This community evolution process is completely transparent to the user. In our framework, community membership of each node is continuously evolving. This allows users to move between communities as their interest shifts between topics, thus enhancing search performance for users all the time even when their interest changes. It also gives our framework great flexibility as it allows communities to dissolve and new communities to form and evolve over time to reflect the latest user interests. In addition to self-sustainability and performance enhancements, our framework has the potential of building extremely large collections although every node is only maintaining a small collection of digital objects. |
