Research On Discoverying The Structure Of Fragmented UGC

The development of current network environment and the improvement of information technology have changed the traditional mode of network content generation and propagation,igniting customers' passion of creation to a great extent.The success of online communities,such as Weibo,Zhihu etc,allowed people to enjoy the convenience of UGC(User-generated content).It contains a wealth of experiences and knacks that people summarized from their work and life,and become an important source of knowledge.However,UGC,in a form of fragmentation,is disordered,multilateral distributed,heterogenous and incomplete,these features make it hard to be used directly and valued correctly.Thus,it is significant and difficult problem that how to find an efficient method to organize fragmented UGC knowledge in the field.This paper,from the point view of social network analysis,designs the discovery model of fragmented UGC knowledge structure,which includes identifying knowledge nodes,building knowledge network and discovering knowledge structure.First of all identifying knowledge nodes includes three steps:document predisposition,unknown word recognition and key phrase extraction.Then according to the intensity of co-occurrence relationship as well as the affiliation between users and the identified knowledge nodes,we build knowledge networks base on interdependence and user cognitive st:ructure.Finally,we apply the technique when selecting core-edge analysis with multilevel factional analysis base on the features of knowledge network as the discovery method.The specific process is as follows,we use the method of statistics of string frequency to recognize unknown words that don't exist in the dictionary,after went through segmented word concatenation,multistage filtering of splicing results and optimized post-treatment,then we extract the key phrase in UGC text as knowledge node.Next,we take advantages of the different relationship between the nodes and build the co-occurrence network of knowledge nodes and affiliation network between users and knowledge nodes,at the same time,we convert this affiliation network to derived knowledge network base on the number of common users.Then we perform core-edge analysis and multilevel factional analysis to these two knowledge networks.In the process of core-edge analysis,we analyzed the discrete core-edge association deletion model and the continuous core-edge model and found the core-edge structure and core-semi-edge-edge structure based on the importance of knowledge nodes;In the process of multilevel factional analysis,we make use of the properties of knowledge content and the definition of factions to classify fine-grained knowledge nodes by each layer,and form a hierarchical knowledge structure based on knowledge content eventually.It achieves the transformation offine-grained knowledge nodes to medium-grained and coarse-grained knowledge content.Finally,in the perspective of users' cognition and degree of knowledge relevance,we compare the results of the two fragmented UGC knowledge structure division and explain the differences between them integrated with the practical reality.In empirical analysis,this paper takes UGC text data under the hashtable function topic section of zhihu,an online knowledge q&a community,as the research object to conduct empirical research to verify the effectiveness of the model,and finally obtains clear experimental results.This paper also discusses their application scenarios and concludes that the hierarchical structure in co-occurrence knowledge network is conducive to the combination of knowledge and the realization of knowledge increment.The hierarchical structure of derived network is more in line with people's cognitive rules of knowledge,and can help people more effectively obtain the required knowledge content,construct knowledge system and optimize the knowledge structure.