Debunking Rumors on Twitter with Tree Transformer

Ma, Jing; Gao, Wei

doi:10.18653/v1/2020.coling-main.476

Cited by 40 publications

(19 citation statements)

References 26 publications

(38 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kumar and Carley (2019) replaced CRFs with Binarised Constituency Tree LSTMs, and used pre-trained embeddings to encode the tweets. More recently, Tree (Ma and Gao, 2020) and Hierarchical (Yu et al, 2020) Transformers were proposed, which combine post-and threadlevel representations for rumour debunking, improving previous results on RumourEval '17 (Yu et al, 2020). Kochkina et al (2017Kochkina et al ( , 2018 split conversations into branches, modelling each branch with branched-LSTM and hand-crafted features, outperforming other systems at RumourEval '17 on stance detection (43.4 F1).…”

Section: Approachesmentioning

confidence: 92%

A Survey on Stance Detection for Mis- and Disinformation Identification

Hardalov¹,

Arora²,

Nakov³

et al. 2022

Findings of the Association for Computational Linguistics: NAACL 2022

View full text Add to dashboard Cite

Understanding attitudes expressed in texts, also known as stance detection, plays an important role in systems for detecting false information online, be it misinformation (unintentionally false) or disinformation (intentionally false information). Stance detection has been framed in different ways, including (a) as a component of fact-checking, rumour detection, and detecting previously fact-checked claims, or (b) as a task in its own right. While there have been prior efforts to contrast stance detection with other related tasks such as argumentation mining and sentiment analysis, there is no existing survey on examining the relationship between stance detection and mis-and disinformation detection. Here, we aim to bridge this gap by reviewing and analysing existing work in this area, with mis-and disinformation in focus, and discussing lessons learnt and future challenges.

show abstract

Section: Approachesmentioning

confidence: 92%

A Survey on Stance Detection for Mis- and Disinformation Identification

Hardalov¹,

Arora²,

Nakov³

et al. 2022

Findings of the Association for Computational Linguistics: NAACL 2022

View full text Add to dashboard Cite

show abstract

“…For example, Dun et al (2021) and Kochkina et al (2018) encoded all news through the Transformer encoder to generate the representation of the news and extract text features for subsequent calculations. Ma and Gao (2020) proposed a Transformer tree-based method to utilize user interaction in conversations. The study models the propagation of each statement as a tree structure.…”

Section: Rumor Detection Based On Transformermentioning

confidence: 99%

Research Status of Deep Learning Methods for Rumor Detection

Tan,

Wang,

Jia

et al. 2022

Preprint

View full text Add to dashboard Cite

To manage the rumors in social media to reduce the harm of rumors in society. Many studies used methods of deep learning to detect rumors in open networks. To comprehensively sort out the research status of rumor detection from multiple perspectives, this paper analyzes the highly focused work from three perspectives: Feature Selection, Model Structure, and Research Methods. From the perspective of feature selection, we divide methods into content feature, social feature, and propagation structure feature of the rumors. Then, this work divides deep learning models of rumor detection into CNN, RNN, GNN, Transformer based on the model structure, which is convenient for comparison. Besides, this work summarizes 30 works into 7 rumor detection methods such as propagation trees, adversarial learning, cross-domain methods, multi-task learning, unsupervised and semi-supervised methods, based knowledge graph, and other methods for the first time. And compare the advantages of different methods to detect rumors. In addition, this review enumerate datasets available and discusses the potential issues and future work to help researchers advance the development of field.

show abstract

“…Considering the close correlations among rumor and stance categories, multi-task learning framework was utilized to mutually reinforce rumor verification and stance detection tasks [20,21,34,46]. To model complex structures of post propagation, more advanced approaches were proposed to use tree kneral-based method [33,47], tree-structured RvNN [35,46], hybrid RNN-CNN model [27], variants of Transformer [16,31], graph coattention networks [28] and Bi-directional Graph Neural Networks (BiGCN) [4,23] for classify different kinds of rumors. Inspired by their success, we develop our approach based on the tree-structured model.…”

Section: Related Workmentioning

confidence: 99%

“…To further capture the complex propagation patterns, kernel learning algorithms were designed to compare different propagation trees [33,41,47]. Propagation trees were also utilized to guide feature learning for classifying different types of rumors on Twitter based on Recursive Neural Networks (RvNN) [35] or Transformer-based models [16,31].…”

Section: Introductionmentioning

confidence: 99%

“…For instance, the voices of opposition and doubt seem to always appear among the posts along with the spread of rumors [36]; also, there are many skepticism and enquiries posts sparked to find out the claim's veracity [57]. Moreover, propagation structure provides valuable clues regarding how a claim is transmitted and opinionated by users over time [31,46]. Figure 1 exemplifies the dissemination of four types of rumorous claims, where each claim is represented as a propagation tree consisting of the source post and user responses.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Weakly Supervised Propagation Model for Rumor Verification and Stance Detection with Multiple Instance Learning

Yang¹,

Ma²,

Lin³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

The diffusion of rumors on social media generally follows a propagation tree structure, which provides valuable clues on how an original message is transmitted and responded by users over time. Recent studies reveal that rumor verification and stance detection are two relevant tasks that can jointly enhance each other despite their differences. For example, rumors can be debunked by crosschecking the stances conveyed by their relevant posts, and stances are also conditioned on the nature of the rumor. However, stance detection typically requires a large training set of labeled stances at post level, which are rare and costly to annotate.Enlightened by Multiple Instance Learning (MIL) scheme, we propose a novel weakly supervised joint learning framework for rumor verification and stance detection which only requires bag-level class labels concerning the rumor's veracity. Specifically, based on the propagation trees of source posts, we convert the two multi-class problems into multiple MIL-based binary classification problems where each binary model is focused on differentiating a target class (of rumor or stance) from the remaining classes. Then, we propose a hierarchical attention mechanism to aggregate the binary predictions, including (1) a bottom-up/top-down tree attention layer to aggregate binary stances into binary veracity; and (2) a discriminative attention layer to aggregate the binary class into finer-grained classes. Extensive experiments conducted on three Twitter-based datasets demonstrate promising performance of our model on both claim-level rumor detection and post-level stance classification compared with state-of-the-art methods. CCS CONCEPTS• Computing methodologies → Natural language processing.

show abstract

Debunking Rumors on Twitter with Tree Transformer

Cited by 40 publications

References 26 publications

A Survey on Stance Detection for Mis- and Disinformation Identification

A Survey on Stance Detection for Mis- and Disinformation Identification

Research Status of Deep Learning Methods for Rumor Detection

A Weakly Supervised Propagation Model for Rumor Verification and Stance Detection with Multiple Instance Learning

Contact Info

Product

Resources

About