Detection of constant member and overlapping community from dynamic literary network

A literary network consists of some characters extracted from any literary text and depicts their relationships, which is a significant example of the social network. The community structures among the members in the narrative are dynamic as the plot reveals from time to time. Some members of the network may reside in more than one community simultaneously and form an overlapping community structure. Some groups of nodes may always stay together in the same community irrespective of time. These nodes are classified as the constant members of the community and behave like authoritative characters of that network for the specified time. Detection of the variable communities within the different classes of people is always challenging for literary researchers. In this paper, we have proposed a dynamic, overlapping community and constant members detection method to study the variability in the social network through the timeline. The study has done by maximizing the stability of all the existing communities based on a various complex graph measuring metrics. As a case study, we have analysed two dramas presented in different languages, namely Strife and Nabanna, written by the Nobel laureate John Galsworthy and renowned play writer Bijon Bhattacharya, respectively, to analyze the varying community structure. We have revealed various features of the characters from the detected communities and compared them with the original literature. Our results confirm a similar social scenario as depicted in the literary domain. The analytical results over the benchmark data sets are more accurate compared to the other state-of-the-art methods and less biased than the human perspective.