Coupling Story to Visualization: Using Textual Analysis as a Bridge - PowerPoint PPT Presentation

Coupling Story to Visualization: Using Textual Analysis as a Bridge Between Data and Interpretation Ronald Metoyer, Qiyu Zhi , Bart Janczuk, Walter Scheirer University of Notre Dame � 1

� 2

� 3

� 4

Narrative Text + Visual Evidence � 5

Narrative Text Visual Evidence � 6 From http://www.espn.com/nba/

Narrative Text Visual Evidence https://www.nytimes.com/interactive/2017/01/30/us/politics/trump-immigration-ban-demographics.html � 7

Integrating narrative text with visual evidence is widely used. � 8

Narrative text and visual evidence are often independently presented. � 9

http://www.espn.com/nba/game?gameId=400975463 � 10

Limitations: Interpret the Interpret the narrative in story from context visualization � 11

Narrative Text Visual Evidence � 12

Deep coupling of narrative text and visualization � 13

Deep coupling of narrative text and visualization � 14

Identifying Story Elements Who When Story Where What � 15

Identifying Story Elements Who When Story Where What � 16

Identifying Story Elements Who When Story Where What � 17

Identifying Story Elements Who When Story Where What � 18

Identifying Story Elements Who When Story Where What � 19

Identifying Story Elements Who When The Warriors trailed by six with three minutes left before Durant , criticized for leaving Oklahoma City last summer to chase a championship, brought them back, scoring 14 in the paint . Where What � 20

Identifying Story Elements: Who Named Entity Recognition (NER) and existing database for team name, player name, coach name, referees in the league � 21

Identifying Story Elements: When and Where Regular Expression and Grammar Rules � 22

Identifying Story Elements: What Labeled sentences from 13 Story Grammar story posts SVM Classifier 98% accuracy on 150 test sentences � 23

Identifying Story Elements: What Labeled sentences from 13 Story Grammar story posts SVM Classifier 98% accuracy on 150 test sentences � 24

Identifying Story Elements: What Labeled sentences from 13 Story Grammar story posts SVM Classifier 98% accuracy on 150 test sentences � 25

Identifying Story Elements: What Labeled sentences from 13 Story Grammar story posts SVM Classifier 98% accuracy on 150 test sentences � 26

Deep coupling of narrative text and visualization � 27

Interaction Design: narrative text to visual element � 28

Interaction Design: visual element to narrative text � 29

Interaction Design: linking visual elements � 30

Case Study Domain expert with 5 years experience as a media producer � 31

Case Study “I love the idea of being able to experience the article, but really dial-in on a specific part of it myself” � 32

Case Study “If the writer emphasized the number of points Curry had from the paint, I’d be interested in seeing that visually, and also exploring for myself where else he may have scored from.” � 33

Case Study “It would be useful if the text highlighted the types of elements - player, stats, times, etc., with different colors” � 34

Implications Create exploratory interaction mechanisms for readers. • More granular analysis • Social media � 35

Implications Reinforce author communication for writers . • Fact checking • Idea generation � 36

Discussion • Generalizability • Evaluation � 37

Conclusion We have presented a novel approach to automatically couple narrative text to visualizations for data-rich stories and demonstrated the approach in the basketball story domain. � 38

Questions? @nickyzhi123 qzhi@nd.edu @ 职启予 Nicky Coupling Story to Visualization: Using Textual Analysis as a Bridge Between Data and Interpretation Ronald Metoyer, Qiyu Zhi , Bart Janczuk, Walter Scheirer University of Notre Dame � 39