Fake news detection on Twitter

• DOI: https://doi.org/10.1108/IJWIS-02-2022-0044
• Published date: 19 September 2022
• Date: 19 September 2022
• Pages: 388-412
• Subject Matter: Information & knowledge management,Information & communications technology,Information systems,Library & information science,Information behaviour & retrieval,Metadata,Internet
• Author: Srishti Sharma,Mala Saraswat,Anil Kumar Dubey

Fake news detection on Twitter

Srishti Sharma

Department of Computer Science and Engineering, The NorthCap University,

Gurugram, Haryana, India

Mala Saraswat

School of Computing Science and Engineering, Galgotias University,

Greater Noida, India, and

Anil Kumar Dubey

ABES Engineering College, Ghaziabad, India

Abstract

Purpose –Owing to the increased accessibility of internet and rela ted technologies, more and more individuals

across the globe now turn to social media for their daily dose of news rather than tradi tional news outlets. With the

global nature of social media and hardly any checks in place on posting of content, exponential increase in spread of

fake news is easy. Businessespropagate fake news to improve their economic standing and influencingcon sumers

and demand, and individuals spread fake news for personal gains like popularity and life goals. The content o ff ake

news is diverse in terms of topics, styles and media platforms, and fake news attempts to distort truth with diverse

linguistic styles while simultaneously mocking true news. All these factors together make fake news detection an

arduous task. This work tried to check the spread of disinformation on Twitter.

Design/methodology/approach –This study carriesout fake news detection using user characteristics

and tweet textual content as features. For categorizing user characteristics, this study uses the XGBoost

algorithm. To classifythe tweet text, this study uses various natural languageprocessing techniques to pre-

process the tweets and then apply a hybrid convolutional neural network–recurrent neural network (CNN-

RNN) and state-of-the-artBidirectional Encoder Representationsfrom Transformers (BERT) transformer.

Findings –This study uses a combinationof machine learning and deep learning approachesfor fake news

detection, namely, XGBoost, hybrid CNN-RNN and BERT. The models have also been evaluated and

comparedwith various baseline models to show that this approacheffectively tackles this problem.

Originality/value –This study proposes a novel framework that exploits news content and social

contexts to learn useful representations for predicting fake news. This model is based on a transformer

architecture,which facilitates representation learning from fake newsdata and helps detect fake news easily.

This study also carries out an investigativestudy on the relative importance of content and social context

features for the task of detecting false news and whether absence of one of these categories of features

hampers the effectiveness of the resultant system. This investigation can go a long way in aiding further

researchon the subject and for fake news detection in the presence of extremely noisy or unusabledata.

Keywords Fake news, Transfer learning, Classification, Transformers, Gradient boosting,

Text classification, Twitter

Paper type Research paper

1. Introduction

The easy accessibility of the internet combined together with the increasing number of social

media channels is perhaps the main reason behind the propagation of fake news at an

unparalleled and unimaginable rate. Content makers across the globe now have new channel s to

The authors thank Kai Shu, Deepak Mahudeswaran, Suhang Wang, Dongwon Lee and Huan Liu

from FakeNewsNet for making their data available, enabling this research.

IJWIS

18,5/6

388

Received28 February 2022

Revised27 June 2022

Accepted1 August 2022

InternationalJournal of Web

InformationSystems

Vol.18 No. 5/6, 2022

pp. 388-412

© Emerald Publishing Limited

1744-0084

DOI 10.1108/IJWIS-02-2022-0044

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1744-0084.htm

fabricate and propagate imaginary articles to build readership or as part of psychological

warfare, financial and political gain. News of questionable credibility erodes the confidence that

everyone has of all news sources, both legitimate and illegitimate. The developing use of

algorithms in robotized news circulation and creation has made it easy and inexpensive to

provide news online at a fast pace. Gartner analysis predicts that “By the year 2022, an

overwhelming number of individuals in technologically advanced countries will devour more

false knowledge than real information”[1]. Social media can be a dangerous outlet if mishandled,

abused or attacked. The spread of any kind of information through social media platforms,

whether false or not, is so quick, that it can lead to grave and irreversible damage in practically no

time. It is palpable that one of the utmost well-known fake news was considerably more rapidly

diffused on a popular social networking giant than the most renowned credible standard

newsflash at the time of ballots in the USA in 2016 (Allcottand Gentzkow, 2017). Approximately

62% people in the USA admitted to getting newscasts via social networks in 2016, whereas in

2012, only 49% testified to obtaining updates via social media. It has also been revealed that

social networking sites on the internet now outrank TV as the significant news source [2].

The issue of spreading disinformation through internet warrants immediate consideration.

Any endeavours to misinform or troll on the internet via spread of disinformation or

disingenuous content sources are now considered grave matters that warrant sincere efforts from

researchers in this field. Therefore, there is an urgent need to develop a system for detecting and

filtering false content. Building such a system is paramount because it can assist both

newsreaders and tech companies. As the dynamic nature and different styles of false news are

major hurdles, we aim to propose a fake news detection scheme that takes into account user

characteristics, content and social conditions. This hybrid approach should provide a robust a nd

effective system for efficiently combating fake news epidemics in the early stages of proliferation.

This article provides anoverview of a fake news detection system that uses tools to detect and

eliminate fake websites from results returned by search giants or news applications. It can be

downloaded and added to the user’s browser or any app that the user uses to get the news feed.

Our major contributionsare outlined as follows:

We propose a novel framework that exploits news content and social contexts to

learn useful representations for predicting fake news. Our model is based on a

transformer architecture, which facilitates representation learning from fake news

data and helps us detect fake news easily. We also use the side information

(metadata) from the news content and the social contexts to support our model to

classify the truth better. Through this work, we propose a fake news architecture

comprising of deep learning and machine learning (ML) (Bidirectional Encoder

Representations from Transformers [BERT] and XGBoost) that can be applied to a

wide range of scenarios, languages and platforms.

We use both news content and social context-based features and conduct an investigative

study into which of these categories of features plays the most important role in detecting

fake news, and if one of these features are unavailable or inadequate in the early stage of

news propagation, whether the effectiveness of the resultant system will be affected.

Many researchers have used multiple categories of features but to the best of our

knowledge, they have not carried out a similar investigation. This investigation can go a

long way in aiding further research on the subject and for fake news detection in the

presence of extremely noisy or unusable data.

In our study, we attempt to develop a model for fake news classification using deep

learning and ML that produces better outcome when compared with the previous

studies.

Fake news

detection

389