Filtering Trolling Comments through Collective Classification

de-la-Peña-Sordo, Jorge; Santos, Igor; Pastor-López, Iker; Bringas, Pablo García

doi:10.1007/978-3-642-38631-2_60

Cited by 3 publications

(6 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Emotions such as anger or rage are clues for detecting a troll's comment [22]. According to the authors of [29,33,34], the information acquired from single comments is not enough to perform a correct analysis and, consequently, they try to integrate methods to verify the consistency of the text according to other comments and their topic. A second research direction involves the Social Network Analysis (SNA) of the communities in order to identify possible trolls [9,28,35] Other analyses on data from users [36,37] are carried out, in order to identify users with antisocial behaviours within a community.…”

Section: Troll Detection Methodsmentioning

confidence: 99%

“…The approach conceived in [33] evaluates the problem from the same point of view, but using different concepts. It is based on the Dempster-Shafer theory, i.e., a generalization of Bayes' probability that turns out to be a very useful tool when it comes to imprecise and uncertain information, like the ones provided by the users of these environments.…”

Section: Thread-based Methodsmentioning

confidence: 99%

“…This approach has the merit of generalizing the study about the trolling phenomenon not only on single comments, but also on its context. Moreover, the framework developed by [33] has not been designed for a particular environment and, since it is generic, it can be applied to any kind of forum. Nevertheless, the article acknowledges that different analysis methods exist to evaluate the nature of the content of the messages and that they are capable to extrapolate the useful data for determining the above-mentioned characteristics.…”

Section: Thread-based Methodsmentioning

confidence: 99%

See 2 more Smart Citations

A Survey on Troll Detection

et al. 2020

View full text Add to dashboard Cite

A troll is usually defined as somebody who provokes and offends people to make them angry, who wants to dominate any discussion or who tries to manipulate people's opinions. The problems caused by such persons have increased with the diffusion of social media. Therefore, on the one hand, press bodies and magazines have begun to address the issue and to write articles about the phenomenon and its related problems while, on the other hand, universities and research centres have begun to study the features characterizing trolls and to look for solutions for their identification. This survey aims at introducing the main researches dedicated to the description of trolls and to the study and experimentation of methods for their detection.

show abstract

Section: Troll Detection Methodsmentioning

confidence: 99%

Section: Thread-based Methodsmentioning

confidence: 99%

Section: Thread-based Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A Survey on Troll Detection

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Early studies of troll detection focus on extracting hand-engineered features from the textual contents of user posts for troll detection [8,10,12,12,25,31]. Signals such as writing style, sentiment as well as emotions have been explored [10,31]. User online activities have also been used to detect trolls [8,12].…”

Section: Troll Detectionmentioning

confidence: 99%

“…• KNN [21]: 𝐾-nearest neighbour classifier with off-the-shelf BERT as the feature extractor. 10 • AdBERT [27] 11 : BERT that fine-tunes an adapter for each campaign. • GPT3 [7] 12 : a very large pretrained model adapted to our tasks using prompt-based learning [24].…”

Section: Datasets and Modelsmentioning

confidence: 99%

MetaTroll: Few-shot Detection of State-Sponsored Trolls with Transformer Adapters

Lin

Zhang

Lau

2023

Proceedings of the ACM Web Conference 2023

View full text Add to dashboard Cite

State-sponsored trolls are the main actors of influence campaigns on social media and automatic troll detection is important to combat misinformation at scale. Existing troll detection models are developed based on training data for known campaigns (e.g. the influence campaign by Russia's Internet Research Agency on the 2016 US Election), and they fall short when dealing with novel campaigns with new targets. We propose MetaTroll, a text-based troll detection model based on the meta-learning framework that enables high portability and parameter-efficient adaptation to new campaigns using only a handful of labelled samples for few-shot transfer. We introduce campaign-specific transformer adapters to MetaTroll to "memorise" campaign-specific knowledge so as to tackle catastrophic forgetting, where a model "forgets" how to detect trolls from older campaigns due to continual adaptation. Our experiments demonstrate that MetaTroll substantially outperforms baselines and state-of-the-art few-shot text classification models. Lastly, we explore simple approaches to extend MetaTroll to multilingual and multimodal detection. Source code for MetaTroll is available at: https://github.com/ltian678/metatroll-code.git CCS CONCEPTS• Computing methodologies → Natural language processing; Information extraction.

show abstract

Deep Learning for Multi-Class Antisocial Behavior Identification From Twitter

Singh

Subramani

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Social Media has become an integral part of our daily life. Not only it enables collaboration and flow of information but has also become an imperative tool for businesses and governments around the world. All this makes a compelling case for everyone to be on some sort of online social media platform. However, this virtuousness is overshadowed by some of its shortcomings. The manifestation of antisocial behaviour online is a growing concern that hinders participation and cultivates numerous social problems. Antisocial behaviour exists in its various forms such as aggression, disregard for safety, lack of remorse, unlawful behaviour, etc. The paper introduces a deep learning-based approach to detect and classify online antisocial behaviour (ASB). The automatic content classification addresses the issue of scalability, which is imperative when dealing with online platforms. A benchmark dataset was created with multi-class annotation under the supervision of a domain expert. Extensive experiments were conducted with multiple deep learning algorithms and their superior results were validated against the results from the traditional machine learning algorithms. Visually enhanced interpretation of the classification process is presented for model and error analyses. Accuracy of up to 99% in class identification was achieved on the ground truth dataset for empirical validation. The study is an evidence of how the cutting-edge deep learning technology can be utilized to solve a real-world problem of curtailing antisocial behaviour, which is a public health threat and a social problem.

show abstract

Filtering Trolling Comments through Collective Classification

Cited by 3 publications

References 7 publications

A Survey on Troll Detection

A Survey on Troll Detection

MetaTroll: Few-shot Detection of State-Sponsored Trolls with Transformer Adapters

Deep Learning for Multi-Class Antisocial Behavior Identification From Twitter

Contact Info

Product

Resources

About