A Hybrid Multitask Learning Framework with a Fire Hawk Optimizer for Arabic Fake News Detection

Mathematics, Год журнала: 2023, Номер 11(2), С. 258 - 258

Опубликована: Янв. 4, 2023

The exponential spread of news and posts related to the COVID-19 pandemic on social media platforms led emergence disinformation phenomenon. phenomenon spreading fake information creates significant concern for public health safety population. In this paper, we propose a detection framework based multi-task learning (MTL) meta-heuristic algorithms in context pandemic. developed uses an MTL pre-trained transformer-based model learn extract contextual feature representations from Arabic posts. extracted are fed alternative selection technique which depends modified version Fire Hawk Optimizer. proposed framework, aims improve rate, was evaluated several datasets experimental results show that can achieve accuracy 59%. It obtained, at best, precision, recall, F-measure 53%, 71%, respectively, all datasets; it outperformed other measures.

Язык: Английский

---

Impacts of COVID-19 pandemic on environment, society, and food security

Environmental Science and Pollution Research, Год журнала: 2023, Номер 30(44), С. 99261 - 99272

Опубликована: Фев. 11, 2023

Язык: Английский

---

The Role of the Crowd in Countering Misinformation: A Case Study of the COVID-19 Infodemic

2021 IEEE International Conference on Big Data (Big Data), Год журнала: 2020, Номер unknown, С. 748 - 757

Опубликована: Дек. 10, 2020

Fact checking by professionals is viewed as a vital defense in the fight against misinformation. While fact important and its impact has been significant, checks could have limited visibility may not reach intended audience, such those deeply embedded polarized communities. Concerned citizens (i.e., crowd), who are users of platforms where misinformation appears, can play crucial role disseminating fact-checking information countering spread To explore if this case, we conduct data-driven study on Twitter platform, focusing tweets related to COVID-19 pandemic, analyzing misinformation, professional checks, crowds response popular misleading claims about COVID-19.In work, curate dataset false statements that seek challenge or refute them. We train classifier create novel 155,468 COVID-19-related tweets, containing 33,237 33,413 refuting arguments. Our findings show volume reach. In contrast, observe surge results quick corresponding increase More importantly, find contrasting differences way crowd refutes some appear be opinions, while others contain concrete evidence, link reputed source. work provides insights into how organically countered social their they amplifying checks. These lead development tools mechanisms empower concerned combating The code data found link. ¹

Язык: Английский

---

g2tmn at Constraint@AAAI2021: Exploiting CT-BERT and Ensembling Learning for COVID-19 Fake News Detection

Communications in computer and information science, Год журнала: 2021, Номер unknown, С. 116 - 127

Опубликована: Янв. 1, 2021

Язык: Английский

---

No Rumours Please! A Multi-Indic-Lingual Approach for COVID Fake-Tweet Detection

Опубликована: Фев. 19, 2021

The sudden widespread menace created by the present global pandemic COVID-19 has had an unprecedented effect on our lives. Man-kind is going through humongous fear and dependence social media like never before. Fear inevitably leads to panic, speculations, spread of misinformation. Many governments have taken measures curb such misinformation for public well being. Besides measures, effective outreach, systems demographically local languages important role play in this effort. Towards this, we propose approach detect fake news about early from media, as tweets, multiple Indic-Languages besides English. In addition, also create annotated dataset Hindi Bengali tweet detection. We a BERT based model augmented with additional relevant features extracted Twitter identify tweets. To expand Indic languages, resort mBERT which fine tuned over Bengali. zero shot learning alleviate data scarcity issue low resource languages. Through rigorous experiments, show that reaches around 89% F-Score detection supercedes state-of-the-art (SOTA) results. Moreover, establish first benchmark two Indic-Languages, Using data, achieves 79% 81% Tweets. Our 78% Tweets without any clearly indicates efficacy approach.

Язык: Английский

---

AraCOVID19-MFH: Arabic COVID-19 Multi-label Fake News & Hate Speech Detection Dataset

Procedia Computer Science, Год журнала: 2021, Номер 189, С. 232 - 241

Опубликована: Янв. 1, 2021

Along with the COVID-19 pandemic, an "infodemic" of false and misleading information has emerged complicated response efforts. Social networking sites such as Facebook Twitter have contributed largely to spread rumors, conspiracy theories, hate, xenophobia, racism, prejudice. To combat fake news, researchers around world are still making considerable efforts build share related research articles, models, datasets. This paper releases "AraCOVID19-MFH"1 a manually annotated multi-label Arabic news hate speech detection dataset. Our dataset contains 10,828 tweets 10 different labels. The labels been designed consider some aspects relevant fact-checking task, tweet's check worthiness, positivity/negativity, factuality. confirm our dataset's practical utility, we used it train evaluate several classification models reported obtained results. Though is mainly for detection, can also be opinion/news classification, dialect identification, many other tasks.

Язык: Английский

---

Overview of the CLEF–2021 CheckThat! Lab on Detecting Check-Worthy Claims, Previously Fact-Checked Claims, and Fake News

Lecture notes in computer science, Год журнала: 2021, Номер unknown, С. 264 - 291

Опубликована: Янв. 1, 2021

Язык: Английский

---

Classification aware neural topic model for COVID-19 disinformation categorisation

PLoS ONE, Год журнала: 2021, Номер 16(2), С. e0247086 - e0247086

Опубликована: Фев. 18, 2021

The explosion of disinformation accompanying the COVID-19 pandemic has overloaded fact-checkers and media worldwide, brought a new major challenge to government responses worldwide. Not only is creating confusion about medical science amongst citizens, but it also amplifying distrust in policy makers governments. To help tackle this, we developed computational methods categorise disinformation. categories could be used for a) focusing fact-checking efforts on most damaging kinds disinformation; b) guiding who are trying deliver effective public health messages counter effectively This paper presents: 1) corpus containing what currently largest available set manually annotated categories; 2) classification-aware neural topic model (CANTM) designed category classification discovery; 3) an extensive analysis with respect time, volume, false type, type origin source.

Язык: Английский

---

The Spread of Propaganda by Coordinated Communities on Social Media

Опубликована: Июнь 24, 2022

Large-scale manipulations on social media have two important characteristics: (i) use of propaganda to influence others, and (ii) adoption coordinated behavior spread it amplify its impact. Despite the connection between them, these characteristics so far been considered in isolation. Here we aim bridge this gap. In particular, analyze interplay with a large Twitter dataset about 2019 UK general election. We first propose evaluate several metrics for measuring Twitter. Then, investigate by different communities that participated online debate. The combination allows us uncover authenticity harmfulness communities. Finally, compare our measures coordination automation (i.e., bot) scores suspensions, revealing interesting trends. From theoretical viewpoint, introduce methodology analyzing dimensions are seldom conjointly considered. practical provide new insights into authentic inauthentic activities during

Язык: Английский

---

Detect Rumors in Microblog Posts for Low-Resource Domains via Adversarial Contrastive Learning

Findings of the Association for Computational Linguistics: NAACL 2022, Год журнала: 2022, Номер unknown

Опубликована: Янв. 1, 2022

Massive false rumors emerging along with breaking news or trending topics severely hinder the truth. Existing rumor detection approaches achieve promising performance on yesterday's news, since there is enough corpus collected from same domain for model training. However, they are poor at detecting about unforeseen events especially those propagated in minority languages due to lack of training data and prior knowledge (i.e., low-resource regimes). In this paper, we propose an adversarial contrastive learning framework detect by adapting features learned well-resourced that low-resourced. Our explicitly overcomes restriction and/or language usage via alignment a novel supervised paradigm. Moreover, develop augmentation mechanism further enhance robustness representation. Extensive experiments conducted two datasets real-world microblog platforms demonstrate our achieves much better than state-of-the-art methods exhibits superior capacity early stages.

Язык: Английский

---