Studying the impact of emotions on the messaging behavior of OSN users on Twitter, Facebook, and YouTube. To this end, an emotion analysis was performed on over 5.6 million social media messages that occurred in 24 systematically chosen real-world events.
4. Introduction
OSNs
Stands for "Online Social Network".
It refers to a platform or website that
enables users to create and share
content, connect with other users,
and interact with each other online.
Examples of OSNs include Facebook,
Twitter, and YouTube.
5. Introduction
Social benefits of OSNs
• Save lives during the 2011 Tsunami disaster in Japan.
• Save lives at Red River flood and Oklahoma fires in
2009.
6. Introduction
Tsunami Disaster
• In March 2011, Japan experienced the most
powerful earthquake and tsunami that caused
massive destruction. While the entire infrastructure
of the region was destroyed, the mobile internet
remained available.
• Resourceful doctors used Twitter to notify
chronically ill patients about where to obtain
essential medicines.
• The tweets spread rapidly through patients'
networks, allowing most of them to attend to their
necessary treatments.
7. Introduction
Red River flood and Oklahoma
fires
• Window of Twitter activity related to the Red River
Flood began on March 8, when residents of the Red
River Valley were operating under threat of flood, and
continued through April 27, when most of the flood
danger had passed.
• For the Oklahoma grassfires ran from April 8 (the day
prior to the grassfire onset), and continued until April 13,
when fire threat ceased.
• The study focused on analyzing the communications
made by individuals who were present during the
Oklahoma Grass fires and the Red River Floods. They
hope this system to improve situational awareness
during emergency events.
8. Introduction
OSNs Negative Influence
on Society
Twitter, Facebook, and YouTube have been used to
spread terrorist propaganda and negatively influence
users (online radicalization).
9. Introduction
Emotions
Refer to the feelings and moods that
individuals experience in response to
various stimuli, such as events, situations,
or interactions with others.
12. Important Terms
Emotional
valence shifts
Negative
events
Sentiment
polarity
Events that potentially
trigger negative emotions
(e.g., war, terror, death).
Refers to the degree of positive,
negative, or neutral sentiment
expressed in a piece of text, such
as a review, tweet, or article.
Refer to changes in the positive
or negative emotional tone of a
person's thoughts, feelings, or
behaviors over time.
Polarizing
events
An situation that divides people into
opposing groups with strongly held
opinions or beliefs (e.g., Political
elections, controversial topics, Social
movements) .
Charged
messages
Refer to messages that
convey strong emotions,
either positive or negative.
Positive
events
Events that potentially
trigger positive emotions
(e.g., birthday celebrations,
festivities).
13. Plutchik’s Wheel of Emotions
• Robert Plutchik was a psychologist
who developed a psychoevolutionary
theory of emotion, considered one of
the most influential classification
approaches for general emotional
responses.
• Plutchik suggested that there were eight
basic emotions; anger, fear, sadness,
disgust, surprise, anticipation, trust
and joy.
14. NRC Emotion Lexicon
• A list of English words and their
associations with eight basic emotions
(anger, fear, anticipation, trust, surprise,
sadness, joy, and disgust) and two
sentiments (negative and positive).
• The annotations were manually done by
crowdsourcing on Mechanical Turk.
• Domain: General
16. Objectives
Presenting a systematic study concerning the
influence of emotional valence shifts on the
messaging behavior of OSN users on Twitter,
Facebook, and YouTube.
Analyzing the intensity of Plutchik’s eight
basic emotions.
Studying polarizing emotion scores
(positive vs. negative).
1
2
3
18. 1 2 3 4
Related
Work
Results
Data
Analysis
Procedure
Discussion
Methodology
19. Related Work
Examined the impact of emojis on message diffusion
patterns.
Extracted emotion polarities for about 19 million
tweets by applying the SentiStrength algorithm.
Analyzed sentiment polarities from tweets related
to the 2010 Winter Olympics.
Conducted a study to examine the role of emotional
valence on the diffusion of anti-tobacco messages.
Assigned sentiment polarities to a data set of about
11,000 tweets by using a Bayesian classifier.
Zhang and
Zhang
Ferrara
and Yang
Gruzd et
al.
Kim et al.
Trung et
al.
Prior studies predominantly examined the impact of
sentiment polarities and emotions on information
diffusion over OSNs.
21. Data Extraction
Published on Twitter, Facebook, and
YouTube.
It’s belong to five different domains
(sports, politics, popular culture, war
and terrorism, and others).
Identified 24 real-world
events
Collected more than 5.6
million messages
Positive, Negative, and Polarizing
events.
Assigned 24 events to one of
the following categories
23. Summarize the events
extracted for each category
N
T Twitter
Number of
messages in
each category
F
Y
Facebook
YouTube
%
Relative size of
each category
26. Data Preprocessing
Raw data cleaning (remove
entries with uninformative
content)
Used Python's langdetect library
for language detection
Remove messages not written in
English
Lemmatization of each
message
Tagging words to their corresponding
part-of-speech category
27. Emotion Extraction
Identifies the presence of 8 basic
emotions by using the NRC word-
emotion lexicon.
Handles negation by changing the valence
of a word (e.g., not happy results in a
negative emotion score for joy: joy = −1)
Deals with intensifiers, downtoners,
and maximizers
Assigns an intensity score for each
emotion in every tweet
Emoticons are identified and categorized
as positive, negative, or conditional
Identifies misspellings and repeated
letters to find “hidden” boosters (e.g., “I
am sooooo happy”)
1
3
4
2
6
5
28. Data Analysis and Research
Questions
• RQ1: Which emotions are expressed during
positive, negative, and polarizing events?
• RQ2: Which messaging behavior do users exhibit
during positive, negative, and polarizing events?
• RQ3: Are there differences in the messaging
behavior when users are faced with messages
that convey expected emotions and those with a
shifted emotional valence?
29. Data Analysis and Research
Questions
• They analyzed how people express
emotions on social media during
different types of events.
• They separated the data into two
groups: one group that expressed
emotions that were expected based on
the type of event, and another group that
expressed emotions that were not
expected. For example, during a positive
event, if people mostly express negative
emotions, that would be considered
unexpected.
• They then analyzed how people behave
in each group, and how they are
affected by expected and unexpected
emotions.
30. Results
Show the Intensities of
Emotions during Positive,
Negative, and Polarizing Events
1
Examine the User behavior as a
reaction to emotionally charged
messages
2
33. Emotion Intensity During Positive,
Negative, and Polarizing Events
Examine whether different emotions
belonging to the same emotional valence
are communicated jointly in a single
message. (e.g., anger with disgust, anger
with joy , trust with joy ).
• Emotions of the same valence are more likely
to be expressed together in a single message
on Twitter due to the character limit, but when
users have more space, positive and negative
emotions are more correlated.
• Emotions of different valences are weakly or
moderately correlated, and different negative
emotions are only weakly or moderately
correlated during positive and polarizing
events.
34. User Behavior
• OSN user behavior is defined as
actions related to sending,
forwarding, and liking public
messages/comments.
• User Behavior on Twitter
• User Behavior on Facebook
• User Behavior on YouTube
35. User Behavior on Twitter
• Positive events trigger more
retweets and likes, users prefer
one-to-one communication during
positive events.
• During negative events users
send more tweets, possibly due
to high arousal emotions
increasing social transmission of
information.
• Polarizing events show the
highest tweeting rate per minute.
SOCIAL
MEDIA
36. User Behavior on Facebook
• Facebook users slightly prefer
replying to and liking Facebook
posts that have a positive
emotion score.
• For negative events, they found
that users tend to reply and
comment predominantly on
negative posts and also send
more messages that convey
negative emotions per day.
• In polarizing events they again
found the highest average
number of comments per unit of
time.
SOCIAL
MEDIA
37. User Behavior on YouTube
• YouTube users tend to "like"
comments on positive events
• The users comply with the base
mood of an event by replying
more to negative messages
during negative events and
"liking" positive messages
during positive events.
• The users reply more to
comments on polarizing events,
unlike Facebook and Twitter
users.
SOCIAL
MEDIA
38. User Behavior on Twitter
• During positive events, there are
fewer tweets conveying
negative emotions, but they still
occur consistently throughout the
extraction period. (Fig. 7a).
• Conversely, during negative
events, there are a considerable
number of tweets conveying
positive emotions, and
unexpected cases where positive
tweets even exceed negative
tweets. (Fig. 7b).
39. User Behavior on Facebook
• (Fig. 8a) shows that positive
emotions dominate over the
negative ones throughout the
entire data extraction period.
• The temporal patterns of expected
and shifted emotions during
negative events resemble those
found on Twitter, with positive
emotions dominating over
negative ones at certain dates,
as observed in (Fig. 8b).
40. User Behavior on YouTube
• In positive events though, we
again predominantly found
messages conveying positive
emotions (Fig. 9a).
• Positive messages dominate
over negative messages on
certain dates during negative
events, similar to findings in
Facebook and Twitter. (Fig. 9b).
41. Discussion
OSN users display positive emotions during negative events
due to social connection needs. (Undoing hypothesis).
Users call for social bonding during tough events
and express vulnerability to trigger compassion.
Users conform to the base emotion of an event by
retweeting, replying, and liking messages that convey
similar emotions.
Emotional messages influence the emotions conveyed in
other users' messages.
43. Future Work
Extend the analysis to studying messages written in
languages other than English.
Investigate how sarcasm is related to shifts in the
emotional valence.
45. Conclusions
Presents a study on the impact of
emotional valence shifts on messaging
behavior of OSN users on Twitter,
Facebook, and YouTube, based on a dataset
of 5.6 million messages from 24 real-world
events.
Analyzes the intensity of Plutchik's eight
basic emotions.
The findings indicate that people prefer
sharing messages that correspond to the
emotional valence of the respective event.
Negative events exhibit predominantly
negative messages but also a surprisingly
large number of positive messages.
46. Photos
● Tsunami Disaster : In wake of Japan disaster,
scientists aim for faster and more accurate
tsunami warnings | The Japan Times
● Red River flood and Oklahoma fires : Texas
And Oklahoma Scorched - Photo 2
(cbsnews.com)
● Emotional Contagion : Everything You Need
To Know About Emotions Contagious
(calmsage.com)
● Yasmeen Izz : Twitter
● Plutchik’s Wheel of Emotions: Plutchik's
Wheel of Emotions: Feelings Wheel • Six
Seconds (6seconds.org)
● Robert Plutchik – Wikipedia
● NRC Emotion Lexicon (saifmohammad.com)
Resources