EventOrient:

A System for Community-Scale Tempo-Contextual Network Analysis of Social Data Streams

Developed in Istanbul Sehir University: Media Lab

• Paper
• Report
• Slides

Abstract

We study the evolution of the social network of Istanbul Şehir University overtime, capturing direct, contextual and latent changes in the network structure. The university's story is embodied in the networks we construct, analyze and temporally monitor. Our system,EventOrient, stands on three components; Web Crawling, Networked Data Analysis and Data storytelling, making it a comprehensive system for community-scale tempo-contextual network analysis. Our goal is to render the social development of the university's community in a lucid and insightful manner.

Data

• The datasets are compiled and published on Datascience Sehir
• A notebook explaining the data used can be found in: datasets

System overview

*EventOrient* provides a thorough pipeline for social network analysis. This pipeline consists of three main components; *Web Crawling*, *Networked Data Analysis* and *Data Storytelling*.

Web Crawling

- The crawler repository can be found in: socialcrawler - Twitter accounts are validated against Gmail contacts obtained from sehir.edu.tr domain

Network Analysis

• Tracking link-formation.

  Constructing a network from twitter connections. This notebook also has the script for filtering twitter accounts to obtain only the ones pertaining to Sehir community.

• Calculating Communities
  Each node is labeled by the community detected by Girvan-Newman algorithm.
  • Affiliation Network Analysis

  Institutional accounts are labled `foci` with which we build an affiliation network. Closures are detected and categorized accross different states of the network in different timestamps
  

Application

• The source code for the Django application of the project can be found in Rest

Backend-Frontend communication

Screenshots