You are on page 1of 28

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/337801955

Detecting Influence in Media with Brexit as Case Study

Preprint · November 2019


DOI: 10.13140/RG.2.2.26384.35843

CITATIONS READS

0 457

1 author:

Sebastian Peterlin
Jožef Stefan Institute
2 PUBLICATIONS 5 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sebastian Peterlin on 07 December 2019.

The user has requested enhancement of the downloaded file.


Detecting Influence in Media with Brexit as Case Study
Sebastian Peterlin et al.
November 15, 2019

Abstract

Relevance & Research Question: This article studies how to detect if and how a news
source is influencing the public’s opinion with the intention to sway an election in one way
or another, by studying Russian news sources in relation to UK news sources leading up
to the 2016 United Kingdom European Union membership referendum, commonly known
as Brexit, as a case study. We hope that our results can be applied to mitigating influence
in future elections.

Methods & Data: In identifying influence, we look for trends and systematic anomalies in 5
forms of media bias specified in Hamborg et al. (2018): event selection, source selection,
commission and omission, labelling and word choice, and size allocation. We used
EventRegistry, a tool which groups articles from news sources in event clusters and
identifies Wikipedia entities (concepts) and their weights in articles (Leban et al. (2014)),
for data on the 6 months leading up to the June 2016 referendum and analysis. We also
used natural language processing tools NLTK and Scikit-learn for analysis.

Results: By identifying pro- and anti- Brexit sources, Wikipedia entities which generally
imply a pro- or anti- Brexit bias, key issues around the election (in the case of Brexit in
2016, migration), and words (and combinations of words) which convey a pro- or anti-
Brexit bias, we find that influence can be detected by identifying trends and systematic
anomalies in 3 forms of media bias: source selection, commission and omission, and
labelling and word choice. Our main finding is that Sputnik and RT, the two major Russian
news sources targeted at an international audience, tended to select UK sources known
to be Eurosceptic and pro-Brexit (pro-Leave), emphasize and add concepts which have a
pro-Brexit bias, de-emphasize concepts which have an anti-Brexit bias, and heavily
emphasize the refugee crisis.

Added Value: By identifying key indicators and by using tools laid out in this article, news
publishers and populations can protect themselves from governments and other
institutions with maligned intentions by flagging articles as providing one-sided coverage
of election topics or warning the public that certain news sources are attempting to
influence an election outcome.

1 Introduction

By studying the 6 months leading up to the 2016 United Kingdom European Union
membership referendum (Brexit), this works seeks to understand if and how Russian
news sources influenced opinion to sway the election in one way or another. Influence,
when used in this context throughout the paper, will be interpreted as bias from news
sources (media bias) with the intention to sway the public’s opinion. Articles having
influence on other articles (or sources having influence on other sources), on the other
hand, when used in this context throughout the paper, will be interpreted as an article (or
source) providing facts, sources, and other information to an article (or source) published
at a later time.

You might also like