A Study Of Online Collaboration Based On Wikipedia

Web 3.0 gives birth to the online collaboration which has been widely accepted.Online knowledge collaborations are prevalent in many open production systems such as Wikipedia,Baidupedia,GitHub and so on.In virtue of these systems,distributed members can collaboratively create a lot of valuable contents.So far,many existing studies from the network science have been brought to analyze the interactive behaviors embedded in these systems.However,how the collaborations influence the qualities of the valuable contents they collaboratively created has not been thoroughly investigated Besides,there are few precedents researching whether the emerging technology such as deep leaining can be utilized to study online collaboration effectively or not.The main work of this thesis as follows.First of all,for the weakness of motif,a model of CoSubGra was proposed in this thesis to research online collaboration.The main process of CoSubGra as follows.Collaboration networks were constructed from Wikipedia dataset to model collaboration at a macro level.Then collaboration subgraphs were mined from them.Finally,based on the characteristics of collaboration subgraphs,how the collaborations among the authors of Wikipedia articles influence the qualities of Wikipedia articles were analyzed from a micro level.The results of experiments as follow.There are many factors influencing the qualities of Wikipedia articles,among which are the collaboration relation among authors of Wikipedia articles,closeness of collaboration among them and the scale of collaboration patterns.A group of authors contribute more to Wikipedia articles than the professional extents of arbitrary individuals in the group.The experiments exhibited the statistical properties of the collaboration subgraphs and the efficiency of them as the metrics for the article quality assessments.Besides,for few previous works applying deep learning to research online collaboration,a model of CoVec was proposed in this thesis to research online collaboration.The main process of CoVec as follows.Co-edit graph was constructed from Wikipedia datasets.Then based on co-edit graph,collaboration relations among authors of Wikipedia articles are encoded into vectors.Finally,the authors were replaced by corresponding vectors to research online collaboration.The results of experiments as follow.The vectors obtained by the model can be utilized to assess article quality effectively.Online collaboration groups were mined from Wikipedia datasets effectively based on these vectors.The high precision can be got from recommending collaborators to some collaborators based on these vectors.