Advanced

Using Twitter to Analyze Swedish Immigration Sentiment : A Comparison between August 2015 and December 2015

Persson, Erik LU (2016) PSYP01 20161
Department of Psychology
Abstract (Swedish)
The aims of this study were to (a) use two automated methods to measure immigration sentiment expressed on Twitter, (b) examine how sentiment changed during the period August 2015 - December 2015, and (c) use 1st person plural pronouns and 1st person singular pronouns as indicators of social and individualistic identity and examine how they moderated sentiment change. All Swedish tweets posted between August 2015 and December 2015 and containing the word flykting (refugee) were collected. Using a manually scored sample, two automated models, one making use of latent semantic analysis and one ignoring latent semantic relationships, were trained. The latter was then used to predict sentiment on a larger sample of tweets. The main analysis... (More)
The aims of this study were to (a) use two automated methods to measure immigration sentiment expressed on Twitter, (b) examine how sentiment changed during the period August 2015 - December 2015, and (c) use 1st person plural pronouns and 1st person singular pronouns as indicators of social and individualistic identity and examine how they moderated sentiment change. All Swedish tweets posted between August 2015 and December 2015 and containing the word flykting (refugee) were collected. Using a manually scored sample, two automated models, one making use of latent semantic analysis and one ignoring latent semantic relationships, were trained. The latter was then used to predict sentiment on a larger sample of tweets. The main analysis consisted of comparing sentiment between August and December and study the moderating effect of pronouns. Both models successfully predicted sentiment. The analysis showed no significant change in overall sentiment but indicated that sentiment diverged. Both 1st person plural pronouns and 1st person singular pronouns were used more frequently among tweeters who expressed positive sentiment, and increased use of 1st person singular pronouns was associated with increased positive sentiment. The proposed methods demonstrate how immigration sentiment expressed on Twitter successfully can be predicted. Complementary methodology is necessary in order to accurately interpret the association between immigration sentiment and the use of pronouns, but the study shows interesting results, strongly encouraging further research on the topic. (Less)
Please use this url to cite or link to this publication:
author
Persson, Erik LU
supervisor
organization
course
PSYP01 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Immigration, refugee crisis, computerized methods, latent semantic analysis, pronouns, social identity theory
language
English
id
8879783
date added to LUP
2016-06-13 09:50:29
date last changed
2016-06-13 09:52:58
@misc{8879783,
  abstract     = {The aims of this study were to (a) use two automated methods to measure immigration sentiment expressed on Twitter, (b) examine how sentiment changed during the period August 2015 - December 2015, and (c) use 1st person plural pronouns and 1st person singular pronouns as indicators of social and individualistic identity and examine how they moderated sentiment change. All Swedish tweets posted between August 2015 and December 2015 and containing the word flykting (refugee) were collected. Using a manually scored sample, two automated models, one making use of latent semantic analysis and one ignoring latent semantic relationships, were trained. The latter was then used to predict sentiment on a larger sample of tweets. The main analysis consisted of comparing sentiment between August and December and study the moderating effect of pronouns. Both models successfully predicted sentiment. The analysis showed no significant change in overall sentiment but indicated that sentiment diverged. Both 1st person plural pronouns and 1st person singular pronouns were used more frequently among tweeters who expressed positive sentiment, and increased use of 1st person singular pronouns was associated with increased positive sentiment. The proposed methods demonstrate how immigration sentiment expressed on Twitter successfully can be predicted. Complementary methodology is necessary in order to accurately interpret the association between immigration sentiment and the use of pronouns, but the study shows interesting results, strongly encouraging further research on the topic.},
  author       = {Persson, Erik},
  keyword      = {Immigration,refugee crisis,computerized methods,latent semantic analysis,pronouns,social identity theory},
  language     = {eng},
  note         = {Student Paper},
  title        = {Using Twitter to Analyze Swedish Immigration Sentiment : A Comparison between August 2015 and December 2015},
  year         = {2016},
}