Research On Information Flow-based Personalized Services

Scientific documents and researchers are inseparable parts of scientific research. Keepingwith up-to-date scientific documents and obtaining consistent help from scientific communityare important in scientific research. Some team members' research activities, e.g., repeatedlysearch for papers, lead to low efficiency of teamwork. An efficient and effective documentsharing method can improve the efficiency and competitiveness of research teams. Aprecondition of sharing documents is the identification of users' interests.Personalization means made for or directed or adjusted to a particular individual. On aWeb site, personalization is the process of tailoring pages to individual users' characteristicsor preferences. Personalization is a means of meeting the customer's needs more effectivelyand efficiently, making interactions faster and easier. Personalization should make users moreeffective by helping them reach their goals. This work has information flow as basis, sharesresources in the Knowledge Grid Environment by personalized services. The maincontributions of this work are: Accurate description of user interests is the precondition for a good personalized service.At present, a popular problem for personalization is the lack of user information. Ouruser profile is created from the information flow that is composed of flow as email flow,instant message flow, and so on. In contrast to user's registration, user's rating, user'srelevance feedback and Web log, information flow is a richer and more persistent dataresource of user information. Building user profile from information flow will make thedescription of user interests more accurate. Introduce the community detection into the construction of user profile and propose thegeneral-to-specific description of user interests. We create community profile byanalyzing structure of information flow network and extract personal profile from thecontent of information flow by referring to the community structures of the informationflow network. We find out similar users according to structure analyzing notinformation filled by users. Discuss the shift problem of user interest and adjust the information flow impact on theuser profile by a time function, which makes the user profile adapt to the changes in theuser interests. Find community structures for the self-organized organization by detecting commoninterest communities in large social network by graph analysis. Our communitydetecting algorithm extends the idea of edge betweenness centrality by introducingweight to differentiate the importance of edges. We also put forward a set of new rulesto direct the algorithm to go on or stop.Organize the document by the RSM (Resource Space Model) that manages resources ina classification-based semantic space. It makes the document locating accurately andrapidly. User metadata, community metadata and document metadata make themanagement of user right and document delivering more simply and more accurately.Recommend right person to help user with information flow network as the basis.Unlike the traditional expert/expertise identification systems which use text analysis asbasis, we find the helper candidates according to community structures and rank themby the relationship analysis between them.The proposed approach is a part of the Knowledge Grid e-Science Platform. Comparedwith relevant works, the proposed approach can bring group awareness, effective documentsharing and valuable recommendations in low costs.