Relevant Word Determination from Dynamic Text Network

Abstract

Text network has been of research interest in the computational domain. A text network contains textual documents with dynamic content. The textual documents linguistically depend upon the grammar and use of words. We have proposed determining relevant words in dynamic content using dynamic complex network approaches. We considered dynamic news updates for exploring our analysis of text networks. This work proposes a complex network-based approach for determining the change of important words with dynamic content. We have suggested determining relevant words from dynamic content using degree centrality, closeness centrality and clustering coefficient. We have proposed Affiliation Factor for bond strength determination among the important words. We have considered an instance of live updates in the news for our approach. The news updates on a particular live incident are uploaded dynamically in a thread. We have introduced the Instance Factor for perceiving the relevance of the important words throughout the document, apart from considering the co-occurrence frequency. The approach is free of semantic analysis and, consequently, exempt depending on any corpus. The analytical results show that the varying important words can be efficiently determined from the dynamic content of the text network compared to the other approaches.