Research On Extraction Of Knowledge Network At Level In Folksonomy Based On Power Law Distribution And Fractal

Folksonomy Knowledge Organization Mode(FKOM)proposed by Thomas Vander Wal firstly has been adopted by all kinds of resource websites to organize web resources since 2004.The difference between Folksonomy Mode and other traditional knowledge organization system is that individual users are free to participate in the marking in the the modern and open semantic network environment in FKOM instead of following the rules regulated by the authorities in this field.Therefore,it appears to be of discrete chaos.For this reason,FKOM has become a hot topic in website studies.In present studies,FKOM and related academic research methods of the construction of tag knowledge from the perspective of network thinking has been widely accepted and recognized.Because of the social tagging in FKOM,there are huge amounts of data processing in the associated researches.With massive data analysis,the advantages obtained from the large data thinking coexist with the "low value" problems at the same time.After all,with the open network environment and free social tag,the social tagging in FKOM is filled with a lot of vague,ambiguous or even wrong information.Therefore,some researchers set threshold to filter data by themselves,which ensures the significance and validity of data to some extent.However,there arise other problems as well.Firstly,the threshold setting lacks necessary theoretical guarantee.Secondly,it is indeterminately equivalent between the original data and the extracted data according to the threshold.Finally,it lacks comparability when faced with problems of multiple periods or multiple types.Therefore,the exploration of the methods of network knowledge extraction which can ensures the prominence of data with a solid theoretical foundation,the equivalency of hierarchical knowledge network and the original knowledge network,and the relative comparability has become an urgent issue in the academic circle.The paper employs 5 sets of domain knowledge data in BibSonomy as data source,which was originated from erection and maintenance of knowledge developed by data engineering team at Kassel University,including.It aims to build domain knowledge network,which can analyze the frequency distribution of incidence statistically based on co-occurrence relationship with tag.With the application of the theory of power-law distribution and fractal,the threshold will be set and the hierarchical knowledge network will be extracted based on the knowledge offrequency.Considering the proved characteristics of power-law distribution from the perspective of degree distribution and the small-world effect in the domain knowledge network based on the co-occurrence of tags,the hierarchical knowledge network extracted in the paper will be tested from two aspects: the power law distribution in degree and the small world effect.The result shows that a hierarchical knowledge network whose threshold has been extracted by knowledge correlation frequency has significant characteristics of power law distribution(scale-free networks)and small-world effect.And based on the result,the equivalence has been proved between hierarchical knowledge network and original knowledge network.Therefore,a hierarchical knowledge network whose threshold has been extracted by knowledge correlation frequency satisfies the characteristics of original knowledge network entirely in FKOM.