Measuring, Understanding, and Classifying News Media Sympathy on Twitter after Crisis Events
Type
conference paper
Date Issued
2018
Author(s)
Abstract
This paper investigates bias in coverage between Western and Arab media on Twitter after the November 2015 Beirut and Paris terror attacks. Using two Twitter datasets covering each attack, we investigate how Western and Arab media differed in coverage bias, sympathy bias, and resulting information propagation. We crowdsourced sympathy and sentiment labels for 2,390 tweets across four languages (English, Arabic, French, German), built a regression model to characterize sympathy, and thereafter trained a deep convolutional neural network to predict sympathy. Key findings show: (a) both events were disproportionately covered (b) Western media exhibited less sympathy, where each media coverage was more sympathetic towards the country affected in their respective region (c) Sympathy predictions supported ground truth analysis that Western media was less sympathetic than Arab media (d) Sympathetic tweets do not spread any further. We discuss our results in light of global news flow, Twitter affordances, and public perception impact.
Keywords
H.5.3. Group and Organization Interfaces: Web-based interaction Twitter
sympathy
sentiment analysis
news media bias
crisis informatics
cross-cultural
crowdsourcing
NLP
File(s)![Thumbnail Image]()
open.access
Name
1801.05802v2.pdf
Size
388.31 KB
Format
Adobe PDF
Checksum (MD5)
dc488a7ccd577bfdda4a3f35cc1e4b6a