Research On Finding Domain Expert And Authoritive Resourse In Social Bookmarking System

Social bookmarking allow users to add free-text labels, or "tags" to Web resource in a shared Web environment and this annotation process become a way of resource organization. Social bookmarking system has get more and more attention due to the ease of use, low cognitive load, lack of controlled vocabulary. However, this kind of feature has brought many problems at the same time. As tags do not restricted by traditional classification, it is unclear how many valuable resources are untagged or tagged with irrelevant tags. In addition, annotation spammers add noise to social bookmarking system. To solve these problems, we put forward a kind of domain expert and authoritative resource discovery model in social bookmarking system based on network topology. The main research content of this article has the following three points:(1) Define the users who bookmark authoritative resource with a given topic before others as domain expert, define a kind of valuable source of information with a given topic as authoritative resource. Then we describe the mutually reinforcing relationship between domain expert and authoritative resource.(2) In the first phase of DEARL, to reduce noise in the Delicious data, we isolate a smaller sub-network, which is "candidate experts". These users’tagging behavior shows potential domain and classificatory expertise. In the second phase, we apply a HITS-based graph analysis to the candidate experts’data to analyze the ranking of the top experts and the topic of authoritative documents. This research propose a distributed method to find the power set of bookmark tag sets, and identify the frequently co-occurring tags shared by many candidate experts, thus we can final confirm the topics of interest in Delicious. Compared with other algorithms, DEARL can accurately find the domain expert and more authoritative resources and extract the related topics of interest. This research also assumed that the process of filtering candidate expert can effectively reduce noise in the Delicious data graph, and can locate the domain experts and authoritative resources very accurately.(3) In evaluation of users’interest topic, three kinds of algorithms that based on HITS actually has no differs. Simultaneously, the candidate expert filtering procedure had no effect on domain expert rankings. However, when we calculated the high-quality candidate expert scores, we find DEARL is the most accurate among the three algorithms in ranking the domain expert.