PRESIDENTIAL CAMPAIGN Josemar Alves Caetano, Jussara Almeida, - - PowerPoint PPT Presentation

▶
presidential campaign
SMART_READER_LITE
LIVE PREVIEW

PRESIDENTIAL CAMPAIGN Josemar Alves Caetano, Jussara Almeida, - - PowerPoint PPT Presentation

Oct 29, 2022 285 likes •860 views

CHARACTERIZING POLITICALLY ENGAGED USERS' BEHAVIOR DURING THE 2016 US PRESIDENTIAL CAMPAIGN Josemar Alves Caetano, Jussara Almeida, Humberto Torres Marques-Neto ASONAM 2018 2 Social networks and political campaings ASONAM 2018 3 Reach of

slide-1
SLIDE 1

CHARACTERIZING POLITICALLY ENGAGED USERS' BEHAVIOR DURING THE 2016 US PRESIDENTIAL CAMPAIGN

Josemar Alves Caetano, Jussara Almeida, Humberto Torres Marques-Neto

slide-2
SLIDE 2

Social networks and political campaings

ASONAM 2018

2

slide-3
SLIDE 3

Reach of the candidates on Twitter (election day)

17 million followers 35 thousand published tweets 12 million followers 9 thousand published tweets

ASONAM 2018

3

slide-4
SLIDE 4

ASONAM 2018

4

Political biases on social networks

slide-5
SLIDE 5

Advocates

ASONAM 2018

5

Political biases on social networks

slide-6
SLIDE 6

Other political groups

Political Bots Regular Users

ASONAM 2018

6

slide-7
SLIDE 7

Characterize users in an online social network taking into account political biases and therefore different behaviors

ASONAM 2018

7

Main objective

slide-8
SLIDE 8

Characterizations

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group

ASONAM 2018

8

slide-9
SLIDE 9

Feature characterization

ASONAM 2018

9

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group
slide-10
SLIDE 10

Language characterization

ASONAM 2018

10

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group
slide-11
SLIDE 11

Profile characterization

ASONAM 2018

11

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group
slide-12
SLIDE 12

Mood characterization

ASONAM 2018

12

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group
slide-13
SLIDE 13

To perform these characterizations…

ASONAM 2018

13

Which features highlight each group Language Patterns Analysis Mood Variation Analysis Popular users

  • f each group
slide-14
SLIDE 14

Methodology

ASONAM 2018

14

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

slide-15
SLIDE 15

Collecting Twitter data

ASONAM 2018

15

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

slide-16
SLIDE 16

Data collection process

ASONAM 2018

16

slide-17
SLIDE 17

Data collection period

  • Data collected over 122 days (August 1st to November 30th 2016)

August 1st Data collection start September 26th First televised debate October 9th Second televised debate October 19th Second televised debate November 8th Election day November 30th Data collection end

ASONAM 2018

17

slide-18
SLIDE 18

Dataset

# of tweets 23 mi # of users 115 k # of relationships 1.8 mi

ASONAM 2018

18

slide-19
SLIDE 19

Identifying political tweets

ASONAM 2018

19

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

slide-20
SLIDE 20

Candidates references considered

Donald Trump Hillary Clinton @realDonaldTrump @HillaryClinton Trump Hillary DT HC

ASONAM 2018

20

slide-21
SLIDE 21

Political hashtags

Donald Trump Hillary Clinton 1 #Trump #ImWithHer 2 #MAGA #NeverTrump 3 #TrumpTrain #Hillary 4 #TrumpPence16 #HillaryClinton 5 #DrainTheSwamp #Hillary2016 6 #tcot #UniteBlue 7 #Trump2016 #VoteBlue 8 #GOP #HillaryBecause 9 #PJNET #OHHillYes 10 #cco #HillYes

ASONAM 2018

21

slide-22
SLIDE 22

Tweet sentiment analysis

ASONAM 2018

22

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

slide-23
SLIDE 23

How sentiment analysis works?

  • SentiStrength tool
  • Dictionary containing emotional words
  • 5
  • 1

1 5

ASONAM 2018

23

slide-24
SLIDE 24

Political sentiment analysis

I love Hillary Clinton and her ideas. I hate Hillary Clinton and her ideas.

ASONAM 2018

24

slide-25
SLIDE 25

Political sentiment analysis problem

I love Hillary Clinton but I hate Donald Trump. I hate Hillary Clinton but I love Donald Trump.

ASONAM 2018

25

slide-26
SLIDE 26

Sentiment Analysis Non-political Tweets Whole text Political Tweets Tweets about

  • ne candidate

Whole text Tweets about both candidates Words associated with each candidate

ASONAM 2018

26

Sentiment analysis approaches

slide-27
SLIDE 27

Sentiment Analysis Non-political Tweets Whole text Political Tweets Tweets about

  • ne candidate

Whole text Tweets about both candidates Words associated with each candidate

ASONAM 2018

27

Non-political tweets

slide-28
SLIDE 28

Sentiment Analysis Non-political Tweets Whole text Political Tweets Tweets about

  • ne candidate

Whole text Tweets about both candidates Words associated with each candidate

ASONAM 2018

28

Political tweets about one candidate

slide-29
SLIDE 29

Sentiment Analysis Non-political Tweets Whole text Political Tweets Tweets about

  • ne candidate

Whole text Tweets about both candidates Words associated with each candidate

ASONAM 2018

29

Political tweets about both candidates

slide-30
SLIDE 30

Identifying words related to candidates

  • Stanford Parser tool
  • Natural language processor

ASONAM 2018

30

slide-31
SLIDE 31

Hillary’s Advocates Trump’s Advocates Political Bots Regular Users

ASONAM 2018

31

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

Identifying politically engaged user groups

slide-32
SLIDE 32

Data mining process

ASONAM 2018

32

Features

slide-33
SLIDE 33

Removing outliers

ASONAM 2018

33

Eliminated users that did not have political tweets

Features

slide-34
SLIDE 34

Identifying political bots

ASONAM 2018

34

BotOrNot

Features

Users with BotOrNot score >= 0.75

slide-35
SLIDE 35

Feature set engineering

ASONAM 2018

35

User Metadata 6 Political Bias 11 Sentiment Analysis 17

44 Features

Syntax 10 Features

slide-36
SLIDE 36

Identifying Regular Users, Trump’s Advocates, and Hillary’s Advocates

ASONAM 2018

36

Silhouette Index Greedy Selection K-Means

1 3 2 Features

slide-37
SLIDE 37

Two steps clustering

Non-political Bots Regular Users Advocates Hillary’s Advocates Trump’s Advocates

ASONAM 2018

37

slide-38
SLIDE 38

Identifying regular users and advocates

Non-political Bots Regular Users Advocates Hillary’s Advocates Trump’s Advocates

ASONAM 2018

38

slide-39
SLIDE 39

Identifying Trump’s Advocates and Hillary’s Advocates

Non-political Bots Regular Users Advocates Hillary’s Advocates Trump’s Advocates

ASONAM 2018

39

slide-40
SLIDE 40

ASONAM 2018

40

Tweet Tweet Tweet 𝑻𝒄 𝒗 Retweet Tweet Tweet Tweet 𝑻𝒃 𝒗

Mood variation analysis

Collecting Twitter Data Identifying Political Tweets Tweet Sentiment Analysis Identifying Politically Engaged User Groups Mood Variation Analysis

slide-41
SLIDE 41
  • 𝑂𝑞𝑣 𝑢1, 𝑢2 : positive tweets total
  • 𝑂𝑜𝑣 𝑢1, 𝑢2 : negative tweets total
  • 𝑇𝑣 𝑢1, 𝑢2 : −1 ≤ 𝑇𝑣 ≤ 1

Subjective Well-Being definition

Graduate Program in Informatics

41

𝑇𝑣 𝑢1, 𝑢2 =

𝑂𝑞𝑣 𝑢1,𝑢2 − 𝑂𝑜𝑣 𝑢1,𝑢2 𝑂𝑞𝑣 𝑢1,𝑢2 + 𝑂𝑜𝑣 𝑢1,𝑢2

slide-42
SLIDE 42
  • 𝑇𝑣 𝑢, 𝑢 + 𝜀 : SWB after retweet
  • 𝑇𝑣 𝑢, 𝑢 − 𝜀 : SWB before retweet
  • ∆𝑇𝑣 values: −2 ≤ ∆𝑇𝑣≤ 2

Mood variation definition

Graduate Program in Informatics

42

∆𝑇𝑣 = 𝑇𝑣 𝑢, 𝑢 + 𝜀 - 𝑇𝑣 𝑢, 𝑢 − 𝜀

slide-43
SLIDE 43

What does it mean?

  • 2
  • 1

1 2

ASONAM 2018

43

slide-44
SLIDE 44

Results

Which features highlight each group Language Patterns Analysis

ASONAM 2018

44

Mood Variation Analysis Popular users of each group

slide-45
SLIDE 45

Clustering Regular Users and Advocates

ASONAM 2018

45

Regular Users (70,290) Advocates (40,003) µ σ µ σ political discourse 0.0871 0.4083 0.4614 1.5802 avg number of political hashtags related to Trump per tweet

  • 0.0005

0.0088

  • 0.0080

0.0297 avg number of political hashtags related to Hillary per tweet

  • 0.0066

0.0141

  • 0.0318

0.0385 positive/negative bias towards Trump 0.0759 0.0617 0.3431 0.1050 positive/negative bias towards Hillary 0.0833 0.4276 0.6592 2.1534

Sillhouette index: 0.81

slide-46
SLIDE 46

Clustering Hillary’s Advocates and Trump’s Advocates

ASONAM 2018

46

Hillary’s Advocates (26,230) Trump’s Advocates (13,733) µ σ µ σ # hashtags in user's description 0.4030 0.1494 0.3516 0.1886 avg number of words per tweet 0.2787 0.1934 0.3429 0.1961 % tweets with some reference to Trump 0.5578 0.2349 0.7702 0.2532 % tweets with some reference to Hillary 0.8355 0.1864 0.6504 0.2624 std of the sentiment score of tweets with some reference to Trump 3.7241 4.8296 7.8192 5.2273 std of the sentiment score of tweets with some reference to Hillary 0.4692 1.4341 0.7009 1.7545

Sillhouette index: 0.72

slide-47
SLIDE 47

ASONAM 2018

47

Hillary’s Advocates Trump’s Advocates Regular Users Political Bots

Language patterns

slide-48
SLIDE 48

Top 5 Hillary’s Advocates

1 2 3 4 5

ASONAM 2018

48

slide-49
SLIDE 49

Top 5 Trump’s Advocates

ASONAM 2018

49

1 2 3 4 5

slide-50
SLIDE 50

Top 5 Political Bots

Twitter suspended the top 10 Political Bots accounts

ASONAM 2018

50

slide-51
SLIDE 51

Top 5 Regular Users

ASONAM 2018

51

1 2 3 4 5

slide-52
SLIDE 52

Mood variation – Hillary’s tweets

ASONAM 2018

52

slide-53
SLIDE 53

Mood variation – Trump’s tweets

ASONAM 2018

53

slide-54
SLIDE 54

Main contributions

ASONAM 2018

54

  • Better understanding of the political engagement of users on online social

networks

  • How candidates may influence their voters using Twitter
  • How users interact with each other
slide-55
SLIDE 55

Future work

ASONAM 2018

55

slide-56
SLIDE 56

Questions?

ASONAM 2018

56

josemar.caetano@sga.pucminas.br