The state of human-centered NLP technology for fact-checking

Das, Anubrata; Liu, Houjiang; Kovatchev, Venelin; Lease, Matthew

doi:10.1016/j.ipm.2022.103219

Cited by 27 publications

(14 citation statements)

References 110 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The results indicate that the perceptions have far-reaching consequences by translating to offline behaviors, creating divisive factions split on partisan views. In particular, recent research highlights the role of opinion amplification in causing extreme polarization on social networks (Lim & Bentley, 2022 ), which indicates a looming possibility of falsehood and polarization amplifying each other (Cinelli et al, 2021a , b ; Das et al, 2023 ; Vicario et al, 2019 ) with the need to tackle falsehood becoming crucial towards breaking this vicious cycle (Das et al, 2023 ). In this regard, the current research prompts IS researchers to view disinformation as a nuanced phenomenon constituting several variants and sheds light on the differing effects of variants on polarization.…”

Section: Discussionmentioning

confidence: 99%

The Polarizing Impact of Political Disinformation and Hate Speech: A Cross-country Configural Narrative

2023

View full text Add to dashboard Cite

Information and communication technologies hold immense potential to enhance our lives and societal well-being. However, digital spaces have also emerged as a fertile ground for fake news campaigns and hate speech, aggravating polarization and posing a threat to societal harmony. Despite the fact that this dark side is acknowledged in the literature, the complexity of polarization as a phenomenon coupled with the socio-technical nature of fake news necessitates a novel approach to unravel its intricacies. In light of this sophistication, the current study employs complexity theory and a configurational approach to investigate the impact of diverse disinformation campaigns and hate speech in polarizing societies across 177 countries through a cross-country investigation. The results demonstrate the definitive role of disinformation and hate speech in polarizing societies. The findings also offer a balanced perspective on internet censorship and social media monitoring as necessary evils to combat the disinformation menace and control polarization, but suggest that such efforts may lend support to a milieu of hate speech that fuels polarization. Implications for theory and practice are discussed.

show abstract

Section: Discussionmentioning

confidence: 99%

The Polarizing Impact of Political Disinformation and Hate Speech: A Cross-country Configural Narrative

2023

View full text Add to dashboard Cite

show abstract

“…Background. Existing fact checking systems (Guo et al, 2022;Das et al, 2023) primarily rely on two components: 1) a document retrieval model, called "retriever", and 2) a veracity prediction model, called "reader". See Figure 2a for an illustration.…”

Section: Preliminariesmentioning

confidence: 99%

“…The retriever views the input claim as a query and returns the top evidence documents that are deemed relevant to the claim-the search is usually performed over a pre-indexed corpus. As the reader, existing studies usually train a classifier over the concatenation of the retrieved documents and the given claim (Das et al, 2023). 5 As stated by Wadden et al (2020) and Guo et al (2022), the veracity prediction step resembles the natural language inference task (NLI).…”

Section: Preliminariesmentioning

confidence: 99%

“…Following existing studies (Guo et al, 2022;Das et al, 2023), our model adopts the pipeline illustrated in Figure 2a. We individually train each component using the labeled data from the source domain and the unlabeled data from the target domain.…”

Section: Problem Statementmentioning

confidence: 99%

“…More specifically, existing automated fact checking systems rely on a pipeline of components to retrieve evidence documents and to infer the final verdict (Guo et al, 2022). As stated by Das et al (2023), coordinating the components within such a pipeline presents an extra challenge.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Semi-Supervised Text Classification via Self-Pretraining

Karisani

2021

Proceedings of the 14th ACM International Conference on Web Search and Data Mining

View full text Add to dashboard Cite

We present a neural semi-supervised learning model termed Self-Pretraining. Our model is inspired by the classic self-training algorithm. However, as opposed to self-training, Self-Pretraining is threshold-free, it can potentially update its belief about previously labeled documents, and can cope with the semantic drift problem. Self-Pretraining is iterative and consists of two classifiers.In each iteration, one classifier draws a random set of unlabeled documents and labels them. This set is used to initialize the second classifier, to be further trained by the set of labeled documents. The algorithm proceeds to the next iteration and the classifiers' roles are reversed. To improve the flow of information across the iterations and also to cope with the semantic drift problem, Self-Pretraining employs an iterative distillation process, transfers hypotheses across the iterations, utilizes a two-stage training model, uses an efficient learning rate schedule, and employs a pseudo-label transformation heuristic. We have evaluated our model in three publicly available social media datasets. Our experiments show that Self-Pretraining outperforms the existing state-of-the-art semisupervised classifiers across multiple settings. Our code is available at https://github.com/p-karisani/self_pretraining.

show abstract