CLASSIFYING FAKE NEWS ARTICLES USING NATURAL LANGUAGE PROCESSING TO IDENTIFY IN-ARTICLE ATTRIBUTION AS A SUPERVISED LEARNING ESTIMATOR

Abstract

There is a growing need for computational tools
that can provide insights on the dependability of
online content due to the prevalence of false
information in widely-accessible media channels
including social media feeds, news blogs, and
online newspapers. In this study, we explore
methods for detecting fabricated news stories in
real time. There are two sides to our help. We
begin by presenting two new datasets for the
fake news detection problem, which together
span seven distinct news domains. We give
many exploratory analyses aimed at discerning
linguistic differences between fake and genuine
news information, and we discuss the collecting,
annotation, and validation procedure in great
detail. We then use the results of these
experiments to develop reliable false news
detectors. Furthermore, we offer evaluations
contrasting machine and human detection of
bogus news.
The news that circulates through social media
networks is a particularly valuable source of
information today. It's easy to see why people
are so drawn to internet-based news: there's very
little effort required, the information is readily
available, and it spreads quickly. Since Twitter
is one of the most widely used real-time news
platforms, it also ranks highly when it comes to
the dissemination of news. In the past, gossip
has been shown to do significant harm by
disseminating false information.