基于时空域联合建模的领域知识演化脉络分析

PDF下载HTML

Title:

Evolutionary path mining of domain knowledge by joint modeling in space-time domain

作者:

金晨, 谢振平, 任立园, 刘渊

1. 江南大学 数字媒体学院, 江苏 无锡 214122;
2. 江苏省媒体设计与软件技术重点实验室, 江苏 无锡 214122

Author(s):

JIN Chen, XIE Zhenping, REN Liyuan, LIU Yuan

1. School of Digital Media, Jiangnan University, Wuxi 214122, China;
2. Jiangsu Key Laboratory of Media Design and Software Technology, Wuxi 214122, China

关键词:

知识演化; 演化路径; 知识网络; 知识系统; 时空域联合; 骨架聚类; 数字媒体知识

Keywords:

knowledge evolution; evolution path; knowledge network; knowledge systems; space-time domain combination; skeleton clustering; digital media knowledge

分类号:

TP181

DOI:

10.11992/tis.201706023

摘要:

同一领域不同知识概念之间存在演化关系，分析演化关系能有效地梳理领域知识的发展脉络，然而网络知识的碎片化、无序性、大规模等特性使得用户很难准确地分析并获取知识之间的这种关系。针对该问题，本文提出一种基于时空域联合建模的领域知识演化脉络分析方法，该方法首先考虑将知识系统以时空域联合知识网络的形式进行表达，随后采用骨架聚类方法提取历年知识网络演化路径，并按知识概念的发展进行演化路径衔接及路径分析。以数字媒体领域知识为例的实验分析表明，该方法能有效提取按年份发展的领域知识演化路径，对于辅助用户进行领域知识的理解与学习，以及个性化推荐具有显著的价值。

Abstract:

In special technology fields, there might be evolutionary relationships between various knowledge concepts, and these evolutionary relationship can be used to depict the developmental venation of the corresponding technology field. However, the characteristics of fragmentation, disorder, and large scale in domain knowledge systems make it difficult for users to accurately identify these knowledge relationships. To address this problem, in this paper, we propose an evolutionary path mining method based on skeleton clustering and the joint modeling of domain knowledge with respect to the space-time correlation. In this method, first we express the knowledge system as a knowledge network with joint space-time correlations, then we adopt the skeleton clustering method to extract the evolutionary path of the knowledge network. In addition, we analyze the connection between the evolutionary paths based on the development of the knowledge concept. An experimental analysis of the digital media domain shows that the proposed method can effectively extract the evolutionary path of domain knowledge, which has significant value for knowledge learning and personalized recommendation.