Excited to share that our SONIC Faculty Affiliate Moses Boudourides has 4 abstracts accepted at #Networks2021! Congratulations to authors Mark McGown, Maryam Khalili, Yasmin Abdelghaffar, and Moses Boudourides.
For more info about Networks 2021, click here.
For more info about our abstracts, please see below:
Networks of Co-Occurring Proper Nouns in Anne Frank’s Diary
Authors: Moses Boudourides and Mark McGown
This work aimed to examine Anne Frank’s ‘The Diary of a Young Girl’ (Bantam Books 1994) from the perspective of network science. Data was first collected from The Internet Archive as a raw text file, and sentiment as well as proper nouns were extracted using the NLTK and spaCy libraries in Python. The relationships of sententially co-occurring proper nouns in the overall text were then compiled and analyzed using the NetworkX library in Python. Community partitioning, assortativity analysis, ego-centric subgraphs, cliques, and centrality indices of the networks of proper nouns were studied in order to discover the structural entanglement and clustering of co-occurrent terms. Communities were discovered amongst these proper nouns that showed correlation with the relationships between them. Significant shifts in sentiment both between community partitions as well as temporally were found. The role of each proper noun was found to correlate with its relative ranking in specific measures of centrality. This analysis suggested that data science and network analysis are adept tools at aggregating meaning from text documents. The findings of this project would be of interest to scholars of the humanities and digital humanities, data scientists studying text mining and text analytics, social scientists and psychologists working on biographical narrative interpretive methods, and historians of WWII intellectual and social events.
Networks of affect in COVID19 positive subreddit posts
Authors: Moses Boudourides and Maryam Khalili
Our aim is to explore from both data- and network-analytic point of view the dynamics of affect developing in posts in the COVID19positive subreddit. We have been collecting threads of discussions on this subreddit from the date that they have started appearing on March 14 until October 2020. From these discussions, using standard techniques of NLP of POS tagging, we were able to extract verbs in stemmed form. Based on the essential principles of linguistics, we were partitioning verbs into four socio-linguistic categories: action, doxastic, emotive and sensory verbs. Our research hypothesis is that the strongly affective features of the discourse developed in social media around the current COVID19 pandemic entail an increasing use of verbs in the categories that we are focusing. Moreover, these verbs appear to co-occur in sentences the sentimental (analytic) score of which tends to increase as the current situation happens to aggravate even more. For this reason, we are studying temporal (longitudinal) networks of sententially co-occurring verbs in the COVID19positive subreddit in order to trace the discursive ways in which feelings, concerns, opinions and emotions are embroiled, discussed and unfold the experiential narrative of the pandemic inside social media.
The Role of Media Susceptibility in a Model of Influence on the Ideology of Actors in Social Networks
Authors: Moses Boudourides and Yasmin Abdelghaffar
We are considering here a model of social influence based on interactions between people and media. In particular, we are considering a bounded-confidence model, which consists of media and non-media actors being nodes of a directed graph such that each node was assigned an ideology, as a real valued number in a given interval. At each time-step, the ideologies of non-media nodes would be updated based on their interactions with other media and non-media nodes. The concept of media susceptibility is introduced, a parameter that controls how much influence a node receives from media and from other non-media nodes, with the extremes being that nodes are either not influenced by media, or not influenced by other non-media nodes. Using the number of steady states as a metric for the polarization of the network, we have demonstrated that the more individuals relied on media, the greater the polarization effect within the network.
Mixing and Segregation of Authors’ Gender, “Culture” and Open Access in Publications on Digital Humanities from 2000 to 2020
Authors: Moses Boudourides
The aim of this contribution is to stress the significance of measuring the inhering structural mixing that exists in heterogeneous networks. I will do this over a bibliographic dataset extracted from the Web of Science (WoS), which consists of publications in the topic of Digital Humanities (DH) from 2000 to 2020. The basic methodological orientation in my study is focusing on temporal (longitudinal) bipartite hypergraphs with vertices being various fields (columns) extracted from a bibliographical dataset. Working on the co-authorship network (but also on other dual networks, like the co-publication network, the co-research-area network and the word-net of keywords in the publications abstracts), I intend to determine the way and the extent of how homophily/assortativity/segregation or mixing/disassortativity/desegragation of certain attributes (like authors’ gender, publications’ research areas and Open Access type) are structuring the patterns of academic publishing on Digital Humanities.
