Using Synchronised Tag Clouds for Browsing Data Collections Alexandre de Spindler, Stefania Leone, Michael Nebeling, Matthias Geel and Moira C. Norrie 1 Tag Clouds Visual summaries Searching, browsing, impression building Features: Size, color, font type etc. can all be used to visualise features of the underlying data collections No. of occurrences Positioning may purely aesthetic, sorted alphabetically or based on relationship between tags 2 1
Tag Clouds as Visual Summaries Profiling Researchers 3 Tag Clouds as Visual Summaries Current work highlighted in red 4 2
Browsing and Searching Only single selections 5 Faceted Browsing Query refinement through facets Dimensions Dimension Facets Facets Donnerstag, 8. Dezember 2011 Stefania Leone, Institute for Information Systems, ETH Zurich, Switzerland 6 3
Synchronised Tag Clouds Combination of tag clouds and faceted browsing Example based on a subset of publications from our research group 7 Synchronised Tag Clouds 8 4
Synchronised Tag Clouds 9 Framework Development of search interfaces based on synchronised tag clouds SQL extension for defining search interface Example Database: 10 5
Web Interface - Components Browser contains breadcrumb bar and several viewers Viewers represent dimensions/entities of the data set A dimension/entity may define a number of views 11 Synchronisation across Viewers Each tag selection restricts the current view and propagates the updates to the other views 1. author.name = «Moira C. Norrie» 2. conference.title = «CAISE» Updates are propagated along relations in the order of selections Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 12 6
Tag Cloud Characteristics Tags have a representation (textual or non-textual) and a weight Adriana Ispas Alexandre de Spindler Christoph Zimmerli Corsin Decurtins Fabrice Matulic Matthias Geel Michael Nebeling Moira Norrie Stefania Leone Tilmann Zaeschke Visualisation Modes Simple list (alphabetically sorted, weight not taken into account) Lined-based tag cloud Aesthetic tag cloud (spiral visualisation) Corsin Decurtins Matthias Geel Adriana Ispas Stefania Leone Fabrice Matulic Michael Nebeling Moira Norrie Alexandre de Spindler Tilmann Zaeschke Christoph Zimmerli Adriana Ispas Alexandre de Spindler Christoph Zimmerli Corsin Decurtins Fabrice Matulic Matthias Geel Michael Nebeling Moira Norrie Stefania Leone Tilmann Zaeschke Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 13 Viewer Definition: Publications CREATE VIEWER PUBLICATION_VIEWER ( Titles" (SELECT title AS tag FROM Publications AS p LIMIT 30) LIST, View defines a tag and an optional count column used to visualise the data as tag cloud Tag: publication title Count: not defined 14 7
Viewer Definitions: Conferences CREATE VIEWER CONFERENCE_VIEWER ( "Name (SELECT c.name AS tag, COUNT (pd.publication_id) AS count FROM Conferences AS c, Published AS pd WHERE c.id = pd.conference_id GROUP BY pd.publication_id ORDER BY c.name ASC) LINE, Year Tag: conference name Count: number of publications at the conference 15 Year view CREATE VIEWER CONFERENCE_VIEWER (, "Year" (SELECT c.year AS tag, FROM Conferences AS c, Published AS pd WHERE c.id = pd.conference_id ORDER BY c.year DESC) LINE, Tag: conference year Count: not defined 16 8
Viewer Definitions: Authors CREATE VIEWER AUTHOR_VIEWER ( Image (SELECT a.image AS tag, COUNT (ap.publication_id) AS count FROM Authors AS a, Authored AS ap WHERE a.id = ap.author_id GROUP BY ap.publication_id) LINE,, Tag: author image Count: number of authored publications 17 Framework Architecture Extended SQL specification as input Language processor parses specification Browser generator creates HTML template SQL View Manager initialises SQL views and executes queries Tag cloud generator outputs HTML according to visualisation mode Final web interface shown in a web browser 18 9
Conclusion Synchronised tag clouds as flexible search interface Supports impression building and navigation at the same time Explore multiple dimensions A framework to develop synchronised tag cloud browsers Configuration of search interfaces for browsing and querying data collections Definition of browser and viewers based on extended SQL syntax Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 19 Reference Alexandre de Spindler, Stefania Leone, Michael Nebeling, Matthias Geel and Moira C. Norrie, Using Synchronised Tag Clouds for Browsing Data Collections, Proc. 23rd Intl. Conf. on Advanced Information Systems Engineering (CAiSE 2011), London, UK, June 2011 20 10
Questions? 21 SQL View Extension CREATE VIEW <view_name> (tag, [count, <other_columns>]) AS (SELECT <column> AS tag [, COUNT (<column>) AS count, <other_columns>] FROM <table_1>[, <other_tables>] [GROUP BY <column>] [ORDER BY <column> ASC DESC] [LIMIT <number>] [RANGE <number>] Main idea: Use SQL views as intermediaries between viewers (tag cloud renderers) and relations Reserved Columns: tag and count A view represents a single data collection to be visualised Builds upon the standard SQL view definition Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 22 11
New SQL Definitions Viewer and Browser CREATE VIEWER <viewer_name> ( "View Name" <view_name> <inner_view_definition> LIST LINE SPIRAL <other_visualisations >, [<other_views>] CREATE BROWSER <browser_name> ( "Viewer Name" <viewer_name>, [<other_viewers>] Viewers can support one or more views Each view is connected with one visualisation mode (LIST, LINE, SPIRAL) Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 23 Conceptional Design Extension of the relational data model One (database) browser is composed of several viewers One viewer manages one or more views and visualises data collections that map to relations Donnerstag, 8. Dezember 2011 Global Information Systems, ETH Zurich 24 12
Viewer Definitions CREATE VIEWER VR_PUB ( Titles" (SELECT title AS tag FROM Publications AS p LIMIT 30) LIST, 25 In ~90% of the cases tag cloud identified as being from own group In ~75% of the cases tag cloud identified as being from a co-worker In ~45% of the cases tag cloud identified as being from other groups In only ~7% of the cases tag cloud identified as being from researcher of another group Augmented tag cloud increased identification rate, but also fault rate increase in self-confidence has both positive and negative effects 26 13