2018
DOI: 10.1145/3140565
|View full text |Cite
|
Sign up to set email alerts
|

Quantifying Controversy on Social Media

Abstract: Which topics spark the most heated debates on social media? Identifying those topics is not only interesting from a societal point of view, but also allows the ltering and aggregation of social media content for disseminating news stories. In this paper, we perform a systematic methodological study of controversy detection by using the content and the network structure of social media.Unlike previous work, rather than study controversy in a single hand-picked topic and use domain-speci c knowledge, we take a g… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1

Citation Types

0
147
0
4

Year Published

2019
2019
2020
2020

Publication Types

Select...
4
3

Relationship

1
6

Authors

Journals

citations
Cited by 242 publications
(151 citation statements)
references
References 38 publications
0
147
0
4
Order By: Relevance
“…In the connection network (CN ), the friends network (CN F R , the accounts the user follows) outperformed the baseline, while the follower network (CN F L ) achieved the lowest average F-score among all classifiers, even when combined with the friends network. This is potentially because of the sparsity of this network, where finding common followers among different users is less likely compared to finding common accounts they might follow, where it is expected to have people of similar stance following common accounts as a part of the homophily phenomena in social media [2,23]. While user's interaction network showed the best overall performance among all feature sets, Table 4, it was interesting to see preference network outperformed all models in two of the five topics when using the binary classifier.…”
Section: Stance Detection Resultsmentioning
confidence: 99%
See 1 more Smart Citation
“…In the connection network (CN ), the friends network (CN F R , the accounts the user follows) outperformed the baseline, while the follower network (CN F L ) achieved the lowest average F-score among all classifiers, even when combined with the friends network. This is potentially because of the sparsity of this network, where finding common followers among different users is less likely compared to finding common accounts they might follow, where it is expected to have people of similar stance following common accounts as a part of the homophily phenomena in social media [2,23]. While user's interaction network showed the best overall performance among all feature sets, Table 4, it was interesting to see preference network outperformed all models in two of the five topics when using the binary classifier.…”
Section: Stance Detection Resultsmentioning
confidence: 99%
“…Our hypothesis is that user's embedded viewpoint in a post is related to the user's identity which could be better modeled by their interactions and connections in the social network. This idea is related to the concept of homophily in which users with same believes tend to have common interests and group together [2,14,23].…”
mentioning
confidence: 99%
“…Edges represent frequently co-purchased books. Books are classified as Liberal (43), Conservative (49), and Neutral (13). Neutral books are randomly assigned to one of the two communities.…”
Section: Methodsmentioning
confidence: 99%
“…Detecting polarization Recently, a significant body of work has emerged that focuses on measures for characterizing polarization in online social media [2,7,13,19,27]. These works consider mainly the structure in social-media interactions and quantify polarization or compute node polarity scores using network-based techniques.…”
Section: Related Workmentioning
confidence: 99%
See 1 more Smart Citation