Hybrid Onion Layered System for the Analysis of Collective Subjectivity in Social Networks

Abstract

This research aims to analyze the Digital Social Networks (DSN) behavior, constructed from the network’s relationships, interactions, and expressions of users’ private states through collective subjectivity. For this purpose, an onion-ring system called COSSOL has been built in a case study for Twitter, following a hybrid approach to integrate Machine Learning classifiers and structural metrics from Computational Linguistics and Computational Sociology disciplines, respectively. The paper designs two experimentation scenarios divided into cases of collective subjectivity analysis for Colombia under different levels of communities’ granularity. The first case validates the system by performing a cointegration test on the metrics of each construct for the onion rings’ communities. The results show that some communities better propagate their subjective expressions against the disclosed topic when they have a higher network density and a common polarity. Moreover, the most stable communities in polarity towards a topic are those whose members are highly connected. Conversely, communities with a higher centrality index in a subset of members do not exhibit stability in collective subjectivity towards a topic disclosed in that community. The second case validates the model with a series of Social Network Analysis (SNA) metrics with a polarity layer to describe the second onion ring subcommunities and their temporal variation through community recalculation. The results show no polar distributions similar to the bimodal ones representing consensus in the values of the common Thinking Acting and Feeling (TAF) forms. In addition, general negative sentiment is identified for the ten most representative nodes of the subcommunities analyzed.