5. What’s the difference? Opinion leaders are hard to identify Much more data and richer information Possibility to track data automatically and analyze them (digital information, ...) very fast (e.g. direct feedback on campaigns) Everybody has a voice – risks and chances for companies Long tail – opinion gathering Choice of words – no strict rules like in press releases, different styles because of different platforms 3 Customer Communication in Twitter
6. Research Questions Goals: getting a deeper understanding about the dynamics of the structures of communication, the participation of the stakeholder and their sentiments in the communication. Are crisis-related issues in twitter discussed (like in the classic media) and are these discussions characterized by peaks and buzzing-stages? Do involved user post higher frequented in peaks than in buzzing stages? Are the postings in the peaks filled with more sentiment-words than in the buzzing stages? Is there a difference between sentiments in Tweets created by power-tweeters (PT) only and the sentiments in Tweets created by all participants of the sample? 4 Customer Communication in Twitter
7. Agenda Motivation and Background Related Work Research Design Summary Research Approach for the further study 5 Customer Communication in Twitter
8. Sentiment in Twitter Messages Sentiment Analysis Sentiment analysis of Tweets: Events in the social, political, cultural and economic sphere do have a significant, immediate and highly specific effect on the various dimensions of public mood (Bollen et al., 2009). Link measures of public opinion derived from polls to sentiment measured from Twitter messages: Sentiment word frequencies in contemporaneous Twitter messages do correlate with several public opinion time series such as surveys on consumer confidence and political opinion over the 2008 to 2009 period (O’Connor et al., 2010). Study of political tweets around the 2009 German federal election: Tweet sentiment (e.g., positive and negative emotions associated with a politician) corresponds closely to voters’ political preferences(Tumasjanet al., 2010). 6 Customer Communication in Twitter
9. Agenda Motivation and Background Related Work Research Design Summary Research Approach for the further study 7 Customer Communication in Twitter
10. Proceeding Objects ofstudy: the Top10 players in theautomotiveindustry Identification of appropriate keywords using classic print media: Identification of keywords by scanning the New York Times over a periode of two weeks, analyzing these articles which are related to one of the carmakers. Structural analysis of the course topics: Observation, analysis and documentation of public communication inTwitter using the keywords found with the help of a software prototype Cleaningupthedata 8 Customer Communication in Twitter
11. Case selection Identificationof an issue The large-scalecarrecall due to a technical fault in the gas pedals and thebreaks Usingthekeyword-combination „recall/-s“, „Toyota“ Implementation ofthe Issue Scanning fortheperiode 13-31 calendarweek: 732.003 Tweets: „Toyota“ 37.232 Tweets: „recall“ und „Toyota“ 9 Customer Communication in Twitter
14. Sentiment Analysis Classifyingthepolarityof a giventextatthedocument, sentence, orfeature/aspectlevel Linguisticdimensions Positive emotions (positive feelings, optimism) Negative emotions (anger, anxiety, sadness) Example: Creatingsentimentprofileforcompanies, partiesoraffiliatedindividuals (e.g., in the form of positive/negative-emotion scales) 12 Customer Communication in Twitter
15. Findings 13 Customer Communication in Twitter Uniform percentage of sentiment words in the discussion A clear tendency of a stronger polarization in peaks
16. 14 Customer Communication in Twitter Findings Isthere a differencebetweensentiments in Tweets createdby power-tweeters (PT) only and thesentiments in Tweets createdby all participantsofthesample?
17. Agenda Motivation and Background Research Approaches Related Work Summary Research Approach for the further study 15 Customer Communication in Twitter
18. Summary Organization-relatedissuesarediscussed in Twitter Usingtheissuescanningkeywordscanidentifytopicsfortrackingdynamics In crisis situations, more individuals participate in the discussion (the contribution per user does not rise) In peak periods, there are clear trends in the discussion to positive or negative sentiments Measures may differ in different types of discussion 16 Customer Communication in Twitter
19. Agenda Motivation and Background Related Work Research Design Summary Research Approaches for the further study 17 Customer Communication in Twitter