Icwsm Workshop Psychology Social Media Gosling Niederhoffer Slideshare
ICWSM 2016 paper presentation, Megha Arora
-
Upload
precog -
Category
Data & Analytics
-
view
1.421 -
download
0
Transcript of ICWSM 2016 paper presentation, Megha Arora
![Page 1: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/1.jpg)
Emotions, Demographics and Sociability in Twitter
Interactions
Kristina Lerman, Megha Arora, Luciano Gallegos, Ponnurangam Kumaraguru (PK), David Garcia
Session V: Users, Opinions and Attitudes 1 @ ICWSM’16
@meghaarora42
19th May 2016
![Page 2: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/2.jpg)
Who am I?
2
• A proud alumna of
• Going to join for MSCS, with a specialization indata science
• Started working with Dr. Lerman at
• Have been a part of for last 3 years
![Page 3: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/3.jpg)
3
![Page 4: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/4.jpg)
MotivationStructure of social interactions shapes individual fitness andwell-being• What factors affect social interactions?
–Cognitive factors: emotions–Socio-economic factors: income, education, demographics
• Can we empirically validate for online social interactions?
*Glyphicons taken from http://www.flaticon.com/4
![Page 5: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/5.jpg)
MotivationStructure of social interactions shapes individual fitness andwell-being• What factors affect social interactions?
–Cognitive factors: emotions–Socio-economic factors: income, education, demographics
• Can we empirically validate for online social interactions?
*Glyphicons taken from http://www.flaticon.com/4
![Page 6: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/6.jpg)
What we need?
Methodology
Results
Conclusions
![Page 7: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/7.jpg)
What We Need
• What does the structure of social interactions look like?• Social Tie strength and Mobility analysis from Twitter
• How happy, how anxious, how powerful people feel?• Emotions
• How much they earn; what are their education levels, theirethnicities?• Socio-economic characteristics
5
![Page 8: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/8.jpg)
What We Need
• What does the structure of social interactions look like?• Social Tie strength and Mobility analysis from Twitter
• How happy, how anxious, how powerful people feel?• Emotions
• How much they earn; what are their education levels, theirethnicities?• Socio-economic characteristics
5
![Page 9: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/9.jpg)
What We Need
• What does the structure of social interactions look like?• Social Tie strength and Mobility analysis from Twitter
• How happy, how anxious, how powerful people feel?• Emotions
• How much they earn; what are their education levels, theirethnicities?• Socio-economic characteristics
5
![Page 10: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/10.jpg)
What we need?
Methodology
Results
Conclusions
![Page 11: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/11.jpg)
Data Collection
• 6M geo-tagged tweets posted from around Los Angeles County
• 340K users
• 2.6M mentions
6
![Page 12: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/12.jpg)
Demographics Data
Obtained socioeconomic characteristics for all LosAngeles County tracts from the 2012 US Census.
Income | Education | Ethnicity
hsp
7
![Page 13: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/13.jpg)
What is a Tract?
Relatively homogeneousunits with respect topopulation characteristics,economic status, andliving conditions.
8
All tracts in the Los Angeles County
![Page 14: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/14.jpg)
Emotion Analysis
-5 -1 +1 +5
Analyze text of tweets to measure emotions
SentiStrength to measure Negativity | Positivity
9
![Page 15: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/15.jpg)
Emotion Analysis
Analyze text of tweets to measure emotions
WKB (Warriner et al.) lexicon to measureValence | Arousal | Dominance
10
![Page 16: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/16.jpg)
Emotion Analysis
Analyze text of tweets to measure emotions
WKB (Warriner et al.) lexicon to measure
Valence | Arousal | Dominance
Quantifies the level of pleasure or pleasantness expressed by a word → Happiness
10
![Page 17: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/17.jpg)
Emotion Analysis
Analyze text of tweets to measure emotions
WKB (Warriner et al.) lexicon to measure
Valence | Arousal | Dominance
Quantifies the level of activity induced by the emotions associated with a word
10
![Page 18: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/18.jpg)
Emotion Analysis
Analyze text of tweets to measure emotions
WKB (Warriner et al.) lexicon to measure
Valence | Arousal | Dominance
Quantifies the level of subjective power associated with an emotional word
10
![Page 19: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/19.jpg)
Social Interactions Analysis
Social Tie Strength of tract i: how much do people there interact with others?
Si : average social tie strength for tract iwj : is the weight of the jth edge (number of times a usermentions another), andki : number of distinct users mentioned in tract i
11
![Page 20: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/20.jpg)
Social Interactions Analysis
Tie Strengthtract1 tract2
Si = 7.33 (Strong Ties) Si = 1.08 (Weak Ties) 12
![Page 21: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/21.jpg)
Mobility Analysis
Spatial Diversity of tract i: how many other places do people tweet from?
ni : number of tracts from which users who tweeted from tract i alsotweeted from, andpij : proportion of tweets posted by these users from tract j
13
![Page 22: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/22.jpg)
What we need?
Methodology
Results
Conclusions
![Page 23: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/23.jpg)
Weaker Ties → more Happiness
-0.36***
14*p < 0.05, **p < 0.01, ***p < 0.001
… also
Stronger ties are
associated with
more Negativity
![Page 24: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/24.jpg)
Weaker Ties → lower Arousal and higher Dominance
15
-0.36*** 0.14***-0.31***
*p < 0.05, **p < 0.01, ***p < 0.001
![Page 25: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/25.jpg)
Weaker Ties → more Education, more Income, fewer Hispanics
16
-0.12*** 0.35***-0.27***
*p < 0.05, **p < 0.01, ***p < 0.001
![Page 26: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/26.jpg)
Higher Mobility → more Happiness
17
0.32***
*p < 0.05, **p < 0.01, ***p < 0.001
![Page 27: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/27.jpg)
Higher Mobility → more Income, more Education, fewer Hispanics
18
0.08** -0.27***0.35***
*p < 0.05, **p < 0.01, ***p < 0.001
![Page 28: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/28.jpg)
What we need?
Methodology
Results
Conclusions
![Page 29: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/29.jpg)
Conclusions
Cognitive factors (emotions) and demographics of placesaffect the quality of online social interactions• Places with better educated, younger and higher-
earning population are associated with weaker socialties and greater mobility (spatial diversity)– These Twitter users express happier, more positive
emotions
19
![Page 30: ICWSM 2016 paper presentation, Megha Arora](https://reader034.fdocuments.net/reader034/viewer/2022050614/58d1e8bd1a28ab51448b5143/html5/thumbnails/30.jpg)
Conclusions
• Places with more Hispanic residents are associated withstronger social ties and lower mobility (spatialdiversity)– People also express less positive, sadder emotions in
these areas
20