The Library
Rumour stance and veracity classification in social media conversations
Tools
Tools
Tools
Kochkina, Elena (2019) Rumour stance and veracity classification in social media conversations. PhD thesis, University of Warwick.
|
PDF
WRAP_Theses_Kochkina_2019.pdf - Submitted Version - Requires a PDF viewer. Download (11Mb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b3494379~S15
Abstract
Social media platforms are popular as sources of news, often delivering updates faster than traditional news outlets. The absence of verification of the posted information leads to wide proliferation of misinformation. The effects of propagation of such false information can have far-reaching consequences on society. Traditional manual verification by fact-checking professionals is not scalable to the amount of misinformation being spread. Therefore there is a need for an automated verification tool that would assist the process of rumour resolution. In this thesis we address the problem of rumour verification in social media conversations from a machine learning perspective.
Rumours that attract a lot of scepticism in the form of questions and denials among the responses are more likely to be proven false later (Zhao et al., 2015). Thus we explore how crowd wisdom in the form of the stance of responses towards a rumour can contribute to an automated rumour verification system. We study the ways of determining the stance of each response in a conversation automatically. We focus on the importance of incorporating conversation structure into stance classification models and also identifying characteristics of supporting, denying, questioning and commenting posts. We follow by proposing several models for rumour veracity classification that incorporate different feature sets, including the stance of the responses, attempting to find the set that would lead to the most accurate models across several datasets. We view the rumour resolution process as a sequence of tasks: rumour detection, tracking, stance classification and, finally, rumour verification. We then study relations between the tasks in the rumour verification pipeline through a joint learning approach, showing its benefits comparing to single-task learning. Finally, we address the issue of transparency of model decisions by incorporating uncertainty estimation methods into rumour verification models. We then conclude and point directions for future research.
| Item Type: | Thesis or Dissertation (PhD) | ||||
|---|---|---|---|---|---|
| Subjects: | B Philosophy. Psychology. Religion > BC Logic H Social Sciences > HM Sociology Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software |
||||
| Library of Congress Subject Headings (LCSH): | Social media, Rumor, Truth, Verification (Logic) | ||||
| Official Date: | 2019 | ||||
| Dates: |
|
||||
| Institution: | University of Warwick | ||||
| Theses Department: | Department of Computer Science | ||||
| Thesis Type: | PhD | ||||
| Publication Status: | Unpublished | ||||
| Supervisor(s)/Advisor: | Liakata, Maria | ||||
| Sponsors: | Bridges Programmes ; University of Warwick. Department of Computer Science | ||||
| Format of File: | |||||
| Extent: | xiii, 201 leaves : charts, illustrations | ||||
| Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
 |
View Item |
Downloads
Downloads per month over past year
