Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr
1 like563 views
This presentation explores public sentiment analysis through related-tags content network from Flickr using NodeXL. Key definitions and processes in extracting and analyzing user-tagged photos and videos are provided, emphasizing folksonomies and social network graphs. The presentation includes discussions on the implications of such analyses for various fields, incorporating methodologies for interpreting tag networks and associated biases.
Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr
- 1. Understanding Public Sentiment: Conducting a Related- Tags Content Network Extraction and Analysis on Flickr Shalin Hai-Jew Kansas State University 2014 National ExtensionTechnology Conference May 2014
- 2. Presentation Overview • This presentation focuses on how to understand public sentiment through a related-tags content network analysis of public Flickr photos and videos. NodeXL is used to conduct data extractions and visualizations of user-tagged Flickr contents and the resulting “noisy” folksonomies.What mental connections may be made about particular issues based on analysis of text- annotated graphs? Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 2
- 4. DefiningTerms • Public sentiment: community attitude (and understanding) • Tag: electronic label (a form of metadata) • Related tags: label which co-occurs with some frequency with another tag (co-occurrence, association) • Folksonomy: informal and inexpert classification system from electronic tags and keywords • Word sense: the gist of a term based on its usage and nuanced understandings (and definitional evocations) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 4
- 5. DefiningTerms(cont.) • Flickr: a digital content-sharing (photos and videos) social media platform • NodeXL: Network Overview, Discovery and Exploration for Excel, an open-source (Ms-PL) and free add-on to Excel (available on Microsoft’s CodePlex) • Data extraction: the drawing out of raw data from a database; a data crawl • Graph: a two-dimensional diagram depicting data • API: application programming interface • Flickr API key and secret: a unique access code for the data extraction through NodeXL (email verified) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 5
- 6. DefiningTerms(cont.) • Social network graph: a 2D or 3D diagram showing social entities and relationships (nodes-links, vertices-edges) • Related tags network graph: the egocentric network of a specified tag (as vertex); a text-based visualization showing entities and inter-relationships between tags (metadata labels / terms) • (Social, content, other) network analysis: study of relations between entities (often expressed as a node-link diagram) • Content network: the representation of relations between content-based entities in a graph Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 6
- 7. DefiningTerms(cont.) • Metadata: information about data often used to enhance archival of that data: understanding of and access to those resources • Data leakage: information released in an unintended or indirect way • Word sense: the gist of a term based on its usage and nuanced understandings (and definitional evocations) • Partition: the segmentation of a graph into separate parts based on similarity clustering (grouping) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 7
- 8. The Process Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 8
- 9. Text-BasedTags at theTag Link on Flickr Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 9
- 10. Sample Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 10
- 11. A Quick “How-to” on Interpreting RelatedTags Graphs • Center-periphery dynamic (and influence) • Large vs. small clusters (and tag frequency) • Clustering around frequency of association and co- occurrence and represented in spatial proximity and color • Social effects of tagging • Structure (relational) and semantic (meaning, definitional) and syntactic (language mechanics) mining Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 11
- 12. Flickr • 10 years old as of Feb. 10, 2014 • 92 million users across 63 countries • 2 million groups • 1 million photos shared a day • Available in 10 languages • Created by Ludicorp and owned now byYahoo, Inc. • Offers a terabyte per user Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 12
- 13. Early Observations? Questions? Affordances • What sorts of information can you know from such related tags networks? • How direct or indirect is this information? • How confident would you be of the results, and why? Constraints • Any early ideas on limits to related tags network analysis? • How accurately may inferences be made about public sentiments and understandings by such related tags word associations? 13 Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr
- 14. Sample RelatedTags Networks (hopefully somewhat related to National Extension interests and within the limits of available Flickr tags) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 14
- 15. YourTurn! • Your table will be assigned several of the following graphs • Find the core related tags search term (sometimes at the center of the graph unless partitions are used) • Identify the main groups and label them in your own words to the best of your ability • Any sense of the public sentiment? Public understandings of the topic? • See any patterns? Anomalies? Anything worth further investigation? • Be ready to share your findings with the group Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 15
- 16. aquaculture Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 16 1
- 17. personal finance Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 17 2
- 18. PTSD Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 18 3
- 19. health Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 19 4
- 20. mortgage Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 20 5
- 22. safety Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 22 7
- 23. lawn Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 23 8
- 24. forest Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 24 9
- 25. food Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 25 10
- 26. county fair Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 26 11a
- 27. County fair Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 27 11b
- 28. family Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 28 12
- 29. garden Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 29 13
- 30. agriculture Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 30 14
- 31. entomology Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 31 15
- 32. home Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 32 16
- 33. exercise Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 33 17
- 34. community Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 34 18
- 35. horticulture Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 35 19
- 36. farming Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 36 20a
- 37. farming Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 37 20b
- 38. parenting Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 38 21
- 39. pest Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 39 22
- 40. livestock Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 40 23
- 41. craft Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 41 24
- 42. disability Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 42 25a
- 43. disability Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 43 25b
- 44. Think of the Possibilities with… • Generic terms • Controversial terms • Competing terms • Multiple languages • Public individuals • Project names • New scientific terms • Cultural memes • Photo or video contests (elicitations for certain multimedia contents) • Content-based video conversations and video replies Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 44
- 45. A ResearchAngle Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 45
- 46. GeneralWorkflow Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 46
- 47. What May be Asserted? • Apparent patterns • Clusters or groups (textual and visual) • Anomalous connections • “Missing” information (what is not showing up) • Apparent sentiments and attitudes (emotion- and value-laden words) • Apparent implied cultures • Any ideas on how to confirm or disconfirm findings from related tags network analysis? Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 47
- 48. Types of Applied Analyses • Inferences based on evidence and reasoning (induction, deduction) • Emergent pattern analysis • A priori pattern analysis • Term and phrase disambiguation (of unstructured text) • Comparisons and contrasts • Text analyses (frequency counts, word trees, sentiment, others) • Image analyses Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 48
- 49. Text and Image-BasedVersions Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 49
- 50. Re-Visualization in NodeXL • Multi-graph visualizations • Text-based vertices (nodes) • Image-based vertices (nodes) • Labeled links (edges) • Differing layout algorithms (usually Fruchterman-Reingold or Harel- Koren Fast Multiscale) • Dynamic filtering (to control variable range) • Analysis of particular “stand-alone” clusters • Analysis of peripheral nodes / vertices Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 50
- 51. Event-Based RelatedTags Networks • Images related to an event • Video related to an event • The tags related to the event Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 51
- 52. TagText Analysis Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 52 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6 related terms
- 53. Analysis overTime • Changing related tags networks over time • Changing terminology in the tags • Trends and patterns • Term manifestations on different content-sharing platforms (research method transferability) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 53
- 54. Other PotentialVisualizations Outside NodeXL • Tag clouds (word frequency count) • Tag word tree (close related word constructs) • Tag geography (maps of where tags come from) • (These additional visualizations are possible depending on the nature of the dataset and access to text analysis and visualization tools.) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 54
- 55. Using NodeXL for the Related Tags Data Extraction on Flickr A Step-by-Step Walkthrough Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 55
- 56. Starting the Data Crawl • Download and install NodeXL (have a recent version of Excel) • Open NodeXL • Go to NodeXL ribbon • File > Import > From Flickr RelatedTags Network … Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 56
- 57. Defining Parameters of the (RelatedTags) Data Extraction • Fill in the search term (vertex tag) • Define parameters • Select degrees (1 degree = egocentric network / ego neighborhood; 1.5 degrees = transitivity among alters of the ego neighborhood; 2.0 degrees = the ego neighborhoods of the alters) • Adding a sample image from each tag in the network • Fill in the Flickr API key (from Flickr’sThe App Garden) • Click “Okay” Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 57
- 58. Image: Starting the Crawl Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 58 Network Degree
- 59. Image: Saving the Data Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 59 Results of the Data Extraction
- 60. Data Processing • Go to the Analysis section in the ribbon • Select Graph Metrics • Check the boxes next to the graph metrics that you want to extract • Click “Calculate Metrics” • Save Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 60
- 61. Image: Processing the Data Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 61 Graph Metrics (post-processing)
- 62. Image:The Graph MetricsTable Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 62
- 63. Data Processing (cont.) • Identify clusters (groups) by… • In Analysis (in the NodeXL ribbon), under Groups, select the parameters for the grouping • ByVertex Attribute • By Connected Component • By Cluster (select clustering algorithm) • By Motif Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 63
- 64. Image: Identifying Clusters Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 64
- 65. OutputtingVisualizations • Create visualization(s) • In graph pane (at the right), click “Show Graph” • Experiment with graph types • Add imagery to vertices (nodes) • Add details to edges (links) • Change labels in Autofill Columns (underVisual Properties) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 65
- 66. Graph Pane Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 66
- 67. Image: Graph Sampler Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 67
- 68. Image: Graph Sampler Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 68
- 69. Image: Graph Sampler Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 69
- 71. Exporting Imagery • Right click in the image pane to • Copy image to clipboard • Save image to file • Capture screenshot • Save Excel file Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 71
- 72. Time for aWalk-through? • Any terms for our related tags network on Flickr? Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 72
- 73. Caveats to the Uses of Related Tags Network Analysis for Research social computing marketing public relations academic research data journalism Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 73
- 74. Potential Structural Sources of Noise and Error • Limited dataset to certain types of multimedia (created by certain subset of the main population) • Researcher conceptualization and analysis error • Inexpert tagging and noisy data (not fully disambiguated, not mutually exclusive terms, not aligned word forms) • Multilingual data • Incomplete extraction (not false positives, but false negatives) • Ambiguity • Dynamism (changes over time) Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 74
- 75. Some Resources • NodeXL on CodePlex • NodeXL Graph Gallery • Social Media Research Foundation (SMRF) • Flickr • Rodrigues, E.M. & Milic-Frayling, N. (2011). Flickr: Linking people, photos, and tags. Ch. 13. In D.L. Hansen, B. Schneiderman, & M.A. Smith’s Analyzing Social Media Networks with NodeXL: Insights from a ConnectedWorld. 201 – 223. Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 75
- 76. Conclusion and Contact • Dr. Shalin Hai-Jew • Instructional Designer • InformationTechnology Assistance Center • 212 Hale Library • Kansas State University • 785-532-5262 • [email protected] Understanding Public Sentiment: Conducting a Related-Tags Content Network Extraction and Analysis on Flickr 76