Identifying rumours on Twitter
Download as PPTX, PDF2 likes2,043 views
This document presents a research proposal for identifying rumors on Twitter. It first defines rumors as unverified information statements circulating in contexts of ambiguity or threat that help people make sense of situations and manage risk. The proposal then discusses identifying rumors based on their context, function, and content. It suggests detecting sudden increases in word frequencies and mapping word connections to identify rumor-related terms. A rumor annotation scheme is proposed that considers context, function, content and ambiguity levels. Tweets would be categorized as speculation, controversy, misinformation or disinformation using chosen measures. The goal is to better understand rumor spreading online and provide context to interpretations.
![“There are measurable differences in the way
microblog messages propagate. The paper
shows that these differences are related to
the newsworthiness and credibility of the
information conveyed, and describes features
[e.g. number of followers, URLs, sentiment,
etc.] in the tweets, that are effective for
classifying information automatically as
credible or not credible.”
Castillo, C., Mendoza, M. & Poblete, B. (2012). Predicting information credibility in
time-sensitive social media. Internet Research, vol. 23, no. 5, pp. 560-588.
A rumour is transmitted](https://image.slidesharecdn.com/identifyingrumours100414-140415032759-phpapp01/85/Identifying-rumours-on-Twitter-10-320.jpg)
Identifying rumours on Twitter
- 1. Kim Holmberg, PhD (e) [email protected] (w3) http://kimholmberg.fi Research plan/proposal on April 10, 2014 Identifying rumours on Twitter
- 2. ”We define rumor as ’unverified and instrumentally relevant information statements in circulation that arise in contexts of ambiguity, danger or potential threat, and that function to help people make sense and manage risk’ (DiFonzo and Bordia, 2007: 13). This definition highlights three facets of social discourse: The context (the situation and/or psychological need out of which discourse arises), function (what people are trying to accomplish by engaging in the discourse), and content (the types of statements uttered).” DiFonzo, N. & Bordia, P. (2007). Rumor, Gossip and Urban Legends. Diogenes, vol. 54, no. 19. DOI: 10.1177/0392192107073433
- 3. “Rumors arise from situational contexts that are ambiguous, threatening or potentially threatening, and where people feel a psychological need for understanding or security.” • “physical insecurity and desire to enhance their sense of security” • “psychological insecurity and desire to enhance the sense of self” Context of a rumour DiFonzo, N. & Bordia, P. (2007). Rumor, Gossip and Urban Legends. Diogenes, vol. 54, no. 19. DOI: 10.1177/0392192107073433
- 4. Function of a rumour DiFonzo, N. & Bordia, P. (2007). Rumor, Gossip and Urban Legends. Diogenes, vol. 54, no. 19. DOI: 10.1177/0392192107073433 Rumors function to make collective sense (sense-making) in an ambiguous situation and/or manage a threat • “Opportunity to actively avoid or neutralize the threat” • “Interpreting the dreaded negative event in a way that reduces its emotional impact” • “To build oneself up by putting other groups down” • “Rumors can perform other functions, such as entertainment, communication of group norms, and defining social network power structure and boundaries, but this is not their primary role.”
- 5. Content of a rumour DiFonzo, N. & Bordia, P. (2007). Rumor, Gossip and Urban Legends. Diogenes, vol. 54, no. 19. DOI: 10.1177/0392192107073433 • Rumours are information statements (they are declarative rather than directive or interrogatory) • A rumour is transmitted, the information statements are communicated through people • These information statements are perceived as relatively useful in some way • Rumours are unverified
- 6. How can we identify rumours, verify information and track their origin?
- 7. Context – Function – Content Identifying rumours
- 8. Content of a rumour DiFonzo, N. & Bordia, P. (2007). Rumor, Gossip and Urban Legends. Diogenes, vol. 54, no. 19. DOI: 10.1177/0392192107073433 • Rumours are information statements (they are declarative rather than directive or interrogatory) • A rumour is transmitted, the information statements are communicated through people • These information statements are perceived as relatively useful in some way • Rumours are unverified
- 9. A rumour is transmitted
- 10. “There are measurable differences in the way microblog messages propagate. The paper shows that these differences are related to the newsworthiness and credibility of the information conveyed, and describes features [e.g. number of followers, URLs, sentiment, etc.] in the tweets, that are effective for classifying information automatically as credible or not credible.” Castillo, C., Mendoza, M. & Poblete, B. (2012). Predicting information credibility in time-sensitive social media. Internet Research, vol. 23, no. 5, pp. 560-588. A rumour is transmitted
- 11. A rumour is transmitted Data about the tweet (e.g. number of retweets, URLs), tweeter (e.g. activity on Twitter) and tweeter’s network (number of following/followers)
- 12. Bla bla bla bla bla bla bla ... Bla bla bla bla bla bla bla ... Bla bla bla bla bla bla bla ... Rumours are information statements
- 13. Rumours are information statements Bordia, P. & Difonzo, N. (2004). Problem Solving in Social Interactions on the Internet: Rumor As Social Cognition. Social Psychology Quarterly, vol. 67, no. 33. DOI: 10.1177/019027250406700105 Most prevalent types of statements identified by RIAS (Rumor Interaction Analysis System): • Interrogatory: questions seeking information • Sensemaking: implied attempts at solving a problem, namely whether or not the rumor was true • Providing information: were statements that included material relevant to the rumor under discussion • Disbelief: indicated disbelief a person’s disbelief in the rumor • Digressive: were not directly relevant to the original rumor, as in narrating some experience or seeking information that was not relevant.
- 14. • Interrogatory • Sensemaking • Providing information • Disbelief • Digressive
- 15. Rumours are information statements http://www.pheme.eu/ Four types of phemes (rumorous memes) or online rumours based on their level of harmfullness: • Speculation: the forming of a theory or conjecture without firm evidence • Controversy: prolonged public disagreement or heated discussion • Misinformation: false or inaccurate information that is spread unintentionally • Disinformation: intentionally false or inaccurate information that is spread deliberately
- 16. Identifying rumours: Research proposal • Spike and/or Burst detection o detect sudden increases in word frequencies • Co-word analysis o map connections between frequent words • Network analysis o identify word clusters, central words and outliers, but perhaps also data about the tweeters and their social networks to identify the most likely tweeters to start/spread rumours (see Castillo et al., 2012) To identify relevant words and phrases, that may be unique for rumours:
- 17. Identifying rumours: Research proposal Context Information Rumours
- 18. Identifying rumours: Research proposal • Annotation scheme • Tweet level vs word level annotation • Ideally something in between or a combination • Level of ambiguity • Level of uncertainty/skepticism • Level of sensationalism • Sentiment of tweet (see Castillo et al., 2012) • and more...
- 19. Identifying rumours: Research proposal Annotation scheme: Example Context Function Content (that may indicate a possible rumour) Vague, ambiguous, uncertainty
- 20. 0 10 20 30 40 cat1 cat2 cat3 cat4 cat5 0 10 20 30 40 cat1 cat2 cat3 cat4 cat5 0 10 20 30 40 cat1 cat2 cat3 cat4 cat5 Identifying rumours: Research proposal Using the chosen measures categorize tweets into speculative, controversal, misinformation and disinformation 0 10 20 30 40 cat1 cat2 cat3 cat4 cat5 Speculation Controversy Misinformation Disinformation
- 21. Identifying rumours: Research proposal 0 50 100 150 200 250 300 350 Speculation Controversy Misinformation Disinformation Verified information Unclear Perhaps a tweet can simultaneously score on different measures?
- 22. World Economic Forum Top 10 trends of 2014: 10. The rapid spread of misinformation online “Finally, we should remember that every case of misinformation is unique and should be considered independently, paying attention to the complexities of the ecosystem it circulates within. In terms of interpreting misinformation, human evaluation will remain essential to put information into context, and context is ultimately what this is all about.” http://reports.weforum.org/outlook-14/top-ten-trends-category- page/10-the-rapid-spread-of-misinformation-online/