Extracting Social Network and Character Categorization From Bengali Literature

Abstract

Literature network analysis is an emerging area in the computational research domain. Literature network is a type of social network with various distinct features. The analysis explores significance of human behavior and complex social relationships. The story consists of some characters and creates an interconnected social system. Each character of the literature represents a node and the edge between any two nodes offered the interaction between them. An annotation and a novel character categorization method are developed to extract interactive social network from the Bengali drama. We analyze Raktakarabi and Muktodhara, two renowned Bengali dramas of Rabindranath Tagore. Weighted degree, closeness, and betweenness centrality analyze the correlation among the characters. We propose an edge contribution-based centrality and diversity metric of a node to determine the influence of one character over others. High diverse nodes show low clustering coefficient and vice versa. We propose a novel idea to analyze the characteristics of protagonist and antagonist from the influential nodes based on the complex graph. We also present a game theory-based community detection method that clusters the actors with a high degree of relationship. Evaluation on real-world networks demonstrates the superiority of the proposed method over the other existing algorithms. Interrelationship of the actors within the drama is also shown from the detected communities, as underlying theme of the narrations is identical. The analytical results show that our method efficiently finds the protagonist and antagonist from the literature network. The method is unique, and the analytical results are more accurate and unbiased than the human perspective. Our approach establishes similar results compared with the benchmark analysis available in Tagore's Bengali literature.