文章摘要

滕广青.关联驱动的领域知识群落生长[J].中国图书馆学报,2017,43(3):58~71
关联驱动的领域知识群落生长
Correlation-Driven Domain Knowledge Community Growth
投稿时间:2016-12-26  修订日期:2017-02-16
DOI:
中文关键词: 知识网络  知识关联  知识群落  Folksonomy
英文关键词: Knowledge network  Knowledge correlation  Knowledge community  Folksonomy
基金项目:本文系国家自然科学基金项目“基于网络结构演化的Folksonomy模式中社群知识组织与知识涌现研究”(编号:71473035)和教育部人文社会科学研究规划基金项目“基于后结构主义网络分析的Folksonomy模式中社群知识非线性自组织研究”(编号:14YJA870010)的研究成果之一
作者单位
滕广青 东北师范大学计算机科学与信息技术学院 吉林 长春 130117 
摘要点击次数: 3137
全文下载次数: 1139
中文摘要:
      知识的发展演化一直是图书情报学界重点关注的主题。随着复杂网络理论的复兴,以网络思维探索知识发展过程中结构关系的演化成为学术界的共识。本文以知识间关联关系为基础,对社会化标注模式下Folksonomy知识组织模式中领域知识群落的生长展开研究。基于关联频度提取层次网络,利用k—丛和派系识别知识网络中的松散型与紧密型领域知识群落,从频度、关联、数量、规模、时序多个维度进行交叉复现分析。研究结果表明,领域知识群落的基本生长路径为“关联关系→松散群落→紧密群落”;知识间关联关系的增长是领域知识群落生长过程中数量繁衍和规模扩容的保障;关联关系频度和数量的积累是领域知识群落生长过程中核心凝聚的过滤器。知识群落生长模式与规律的揭示,有助于从知识间的互促互扰关系方面拓展领域知识组织视野并把握知识发展脉络。但本研究在维度的划分粒度方面还有待进一步加强。图3。表4。参考文献36。
英文摘要:

    The evolution of domain knowledge has always been the focus of the library and information academia. Exploring the evolution of structural relationships in the process of knowledge development with network thinking has become the consensus of academia. The evolution process of the domain knowledge communities can be tracked and analyzed in view of the changes of the relationships among knowledge units to reveal the growth pattern and the law of the domain knowledge communities. This study makes an extraction of 2 075 related articles in specific knowledge domains in folksonomy knowledge organization mode with 203 tags and the correlations number 2 953 in total. The time span is from 2005 to 2015 and the timer shaft is divided into 11 units of time-window. The original domain knowledge networks are constructed based on the correlations between tags. Using the frequency of correlation as the threshold,the knowledge networks at level are extracted based on scale-free and fractal theory. The k-plexes and cliques in the knowledge networks at level are calculated separately,and the loose knowledge communities and close-knit knowledge communities are identified by k-plexes and cliques. On this basis,the time series analysis on tags,correlations,k-plexes,cliques in the knowledge networks is executed from the quantitative dimension. The long-term trends in the evolution of domain knowledge communities are identified. The number of tags,the number of correlations,the number of k-plexes,the average scale of k-plexes,the number of cliques,and the average scale of the cliques in the original knowledge networks and the knowledge networks at level in the key time windows are cross-discovered. The influence of knowledge correlation on the growth of domain knowledge community is thus revealed. The results show that the basic growth path of the domain knowledge community is ‘correlation→loose knowledge community→close-knit knowledge community. Knowledge correlation achieves the connection between the knowledge nodes. When the knowledge correlations are rich to a certain extent,they structure the loose knowledge communities identified by k-plexes. With the further enrichment of knowledge correlations,the close-knit knowledge communities identified by cliques emerge. The increase of the correlations between knowledge units is the guarantee of the quantity multiplication and scale expansion in the growth of domain knowledge communities. Both quantitative evolution analysis and cross-discovery show that in the case of the same scale of the knowledge network,as long as the numbers of correlations continue to increase,the correlations will still lead to more loose knowledge communities and close-knit knowledge communities. At the same time,the new correlations have the opportunities to bridge the old k-plexes or cliques into larger k-plex or clique,resulting in a decrease in the number of knowledge communities and an increase of scale. The frequency of correlation and the accumulation of numbers are the filter of core cohesion during the growth of domain knowledge communities. The knowledge network at level filters to get the significant correlations by threshold. Some once-prominent knowledge nodes and correlations fade out of the knowledge community,and the number and scale of the communities are reduced and condensed. In this article,a knowledge network at level extraction method based on scale and fractal,which extends the identification way of knowledge and correlation in knowledge network analysis,is proposed to provide a new way for knowledge networks to process large-scale data. The correlation-driven domain knowledge community growth patterns in this study can capture the most critical factors in the evolution of knowledge communities in the process of knowledge evolution. Although there is no elasticity in time granularity and dimensions of cross-discovery,the revelation of correlation-driven domain knowledge community growth patterns helps to grasp the evolution venation of knowledge community,which has a positive effect on revealing the law of domain knowledge development. 3 figs. 4 tabs. 36 refs.

查看全文   查看/发表评论  下载PDF阅读器