The GMAP Co-word Analysis Method And Its Application In The Ancient Village Literature Analysis

Ancient villages attracted people's attention in 1980 s,but it not until 1990 s people attached importance to them.The ancient villages' literature was about archaeology before 1990 s.In recent years,with the increasing awareness of the protection of traditional culture,the ancient villages have been paid more and more attention.The ancient villages have been paid more and more attention by scholars.The number of documents about ancient villages in CNKI has also increased exponentially with time.So,the exploration of ancient village literature has very important research value.Co-word analysis method which is the classical method in bibliometric is widely used in the field of topic discovery.The traditional co-word analysis concludes three processes: term collection process,co-occurrence frequency statistics and clustering analysis process.The traditional co-word analysis exists some problems such as strong subjectivity,insufficient information,unstable clustering,the unreasonable member of cluster,and lacking of semantic analysis of clusters.As a result,there is a deviation easily when detecting the domain topics.This paper propose a new co-word analysis named GMAP analysis which integrates g-index,mutual information theory,affinity propagation clustering and co-word analysis.Firstly,we use g-index to choose the number of high frequency keywords.Secondly,we use mutual information theory to transform co-occurrence matrix into similarity matrix.Finally,we use affinity propagation clustering algorithm to find the domain topics.In order to track the topics dynamically,we proposed a topic evolution framework based on co-word analysis.The framework which is based on the GMAP method uses kullback-leibler divergence with asymmetric properties to measure the similarity between topics and takes advantage of the ThemeRiver model to achieve visualization.This paper firstly investigated the related domestic and foreign researches on traditional co-word analysis by method of document investigation to find the defects.We achieve the goals of obtaining the cluster center automatically and reasonable semantic analysis of class by improving every process of the traditional co-word analysis.Then the validity of the proposed GMAP method is verified by numerical experiments about ancient villages.The framework we proposed in this paper based on GMAP co-word analysis starts from the results choose the current optimal visualization technology and the similarity measure method of the framework matches the visualization technology.In order to validate the practicability of the topic evolution framework,this paper applies it to Chinese ancient villages domain to detect the topics and track the topic evolution,and analyzes the topic evolution of ancient villages domain respectively from the macro and micro level.