#ANZMAC2014 Twitter Content Analysis Framework: Classification and Coding of Twitter Content
Download as PPTX, PDF3 likes1,069 views
The document outlines an updated framework for classifying Twitter content, developed by Dr. Stephen Dann, which categorizes tweets into five main domains: conversational, news, pass along, social presence, and broadcast. It also includes a detailed discussion of various sub-classifications and examples of tweet types, alongside statistical data on tweet usage across different platforms and timelines. The overarching goal is to analyze and model Twitter usage patterns based on the classification system and the data collected.
#ANZMAC2014 Twitter Content Analysis Framework: Classification and Coding of Twitter Content
- 1. Twitter! TWITTER CONTENT ANALYSIS FRAMEWORK: CLASSIFICATION AND CODING OF TWITTER CONTENT
- 2. Dr Stephen Dann You may remember me from… Dann (2010) Twitter Content Classification Framework That special session last year
- 3. The Plan Announce the update patch for Dann (2.0.15) from Dann (2.0.10) Show some data Ask an awkward question Questions (yours)
- 4. The Patch Notes Dann (2015) Grounded Theory Strikes Again Moved “Broadcast” from “Social Presence” to “Status” Collapsed several sub domains Bringing the focal point of the coding to a 3 level domain Create Communicate Curate
- 5. Words.
- 6. The Five Domains 1. Conversational replies 2. News Messages of factual / external commentary 3. Pass along Existing content shared or Retweeted 4. Social Presence Connectivity via ritual textual behaviour 5. Broadcast Opinions and experience
- 7. There’s a lot of literature Response: Tweet that commence with the @ symbol to indicate a directed response to another Twitter user (Cahill 2009; Ratkiewicz 2010; Steiner 2009; Wilson 2008). Further sub-classification of response is not included. Referrals: Any tweet that would fit as pass-along which is directed to another Twitter user (Efron 2011; Honeycutt and Herring, 2009; Naaman, Boase and Lai, 2010). Here, the purpose of the tweet is to provide Twitter-external content link directly to the respondent, rather than to all followers of the account. Rhetorical Presence (previously action) is the use of the @username to acknowledge physical, social or temporal connection (Cranefield and Yoong, 2009; Efron 2011; Hohl 2009; Jansen et al. 2009). Rhetorical Presence can be attribution of content, acknowledgment of physical co-location or introductions of a Twitter user to another’s timeline followers.
- 8. Conversational Identified by a @statement to address another user Response Any tweets which commence or finish with another user’s name and which do not meet the requirements of the referral category Referral @responses containing URLs or recommendation of other Twitter users. (Excludes RT @user) Rhetorical Presence Activities involving other Twitter users, or tweets that describe the presence of other Twitter users. News Identifiable newsworthy content Journalism Coverage of live events including factual, descriptive recounts or opinion and social commentary Real-Time Event Live discussion of an identifiable event such as a conference, live television or live event collected with or without a consistent #hashtag. Press Release Identifiable announcement of a forthcoming event without URLs to an external source such as timetable announcements, schedules and session start times Sport Identifiable results of sporting events or discussions of sporting performances Transport Traffic, transport, flight, road or rail related announcements including accidents and delays Weather Report of prevailing weather conditions inclusive of extreme weather events and natural disasters
- 9. Pass along Tweets as curation of content Annotated Media Tweets that are captions for media hosted on Twitter Curation Posting of third party content for followers via the Twitter URL (t.co) or other URL. Offline source Tweet that contains a reference in APA, Oxford or Harvard format, or a statement in inverted commas to denote a quotation from a third party, speaker or source material Retweets Partial or full reproduction of another tweet marked with “RT”, “retweet”, “MT” or “modified tweet” Social Presence Messages of connected presence Ceremonial Greetings Tweets where the community is addressed indirectly as a whole with the greeting or statements of gratitude Fourth wall Textual equivalent of comments made directly to the camera for an imagined audience Self-referential commentary Tweets directed by the author to themselves through “Note to self” “FYI” or “Just for the record” and function as thought bubble style comments Unclassifiable Catch-all category for cat-on-keyboard input and unclassifiable strings of text
- 10. Broadcast Tweets that express the account holder's experiences Action The diary of daily life tweets which answer “What are you doing?” Reflective Statements that address cognitive or emotive responses that answer “What am I thinking?” or “What am I feeling?” Experience Tweets that relay the physical experience as an answer to “What am I experiencing?” – includes location, physical sensations, temporal experience and interaction Statement Observation of life, stated opinions and streams of consciousness “What do I want the world to know?” and “What are my thoughts on a specific topic?”
- 11. Not just classifications Via Source of the Tweet 96 different software clients (thus far) Timestamp Day of the Week Time of day Office hours | morning / evening
- 12. Via (WA Reelection Candidates) ID Platform N % 63 Web 901 35.0% 54 Twitter for iPhone 668 26.0% 51 Twitter for Android 469 18.2% 6 Echofon 91 3.5% 40 Tweet Button 86 3.3% 42 Tweetbot for iOS 66 2.6% 7 Facebook 55 2.1% 53 Twitter for iPad 51 2.0% 16 iOS 45 1.8% 10 Google 42 1.6% 15 Instagram on iOS 38 1.5%
- 13. Sample Data (WA Reelection Candidates) Preelection Block Election Block Post Election1 Post Election 2 Conversational 615 954 461 743 News 50 97 58 69 Pass along 884 1444 701 899 Social Presence 5 13 4 12 Broadcast 38 63 36 53 1592 2571 1260 1776
- 14. Sample Data Pre-election Election Block Post Election1 Post Election 2 n Presence of Others Conversational - Response 394 643 367 612 2016 Conversational - Rhetorical Presence 60 75 40 56 231 Pass along - Retweets 694 1092 465 617 2868 1148 1810 872 1285 5115 External Content Conversational - Referral 161 236 54 75 526 Pass along - Curation 189 351 231 282 1053 350 587 285 357 1579 Ratio: Internal to External 3.28: 1 3.08: 1 3.05: 1 3.60: 1 3.24: 1
- 15. Monday Tuesday Wednesday Thursday Friday Saturday Sunday N Web 86 172 158 133 174 123 55 901 Twitter for iPhone 95 70 66 135 94 128 80 668 Twitter for Android 49 81 74 42 92 75 56 469 Echofon 12 21 18 3 4 3 30 91 Tweet Button 11 12 16 13 10 12 12 86 Tweetbot for iOS 11 7 3 5 7 22 11 66 Facebook 3 20 22 2 3 4 1 55 Twitter for iPad 7 6 6 12 8 6 6 51 iOS 3 9 8 7 7 9 2 45 Google 6 6 11 3 8 2 6 42 Instagram on iOS 3 3 3 7 7 9 6 38 TweetCaster for Android 1 2 1 3 4 5 14 30 TweetDeck 0 0 3 4 5 0 0 12 Other 2 5 1 3 1 3 2 17 289 414 390 372 424 401 281 2571
- 16. The Question Expressed preference choice data Content | Curate | Communicate Time series Hours | Days | Time blocks Device specific data mobile | desktop | other Where’s the formula?
- 17. The (in)tractable problem If Twitter use is an expressed choice, and we have data that records that expressed choice, can we model and predict? Can we extract the necessary calculations?