1. Concept Relationship Editor: A visual interface to support the assertion of synonymy relationships between taxonomic classifications Paul Craig & Jessie Kennedy Napier University Centre for Information & Software Systems

2.  Taxonomic concepts  Named nodes in a specific classification  Apium graveolens L. sec. Bentham 1867  From 10s – 1,000s in tree Taxonomic data Apium graveolens L. (celery) order family species Classification by Taxonomist A in 1950 genus

3.  Relationship within classifications  Parent/child Taxonomic data Apium graveolens L. (celery) order family species Classification by Taxonomist A in 1950 genus

4.  Taxonomic Ranks  Named levels in tree  Family, species etc  Between 2 and 15  Ordered  Specific ranks used vary within and between classifications Taxonomic data Apium graveolens L. (celery) order family species Classification by Taxonomist A in 1950 genus

5. Taxonomic Classifications A partial display of one medium sized taxonomic classification. All taxa at Order rank are shown, with one, Gruiformes drilled- down to show it’s sub-tree.

6. Multiple Classifications Apium graveolens L. Classification by Taxonomist A in 1950 Classification by Taxonomist B in 2000 ?

7. Multiple Classifications  Set-based notation symbols used to specify between classification relationships:  ≡ (congruent with)  ≢ (not congruent with)  (contains)  (is contained in)  ⊗ (overlaps with) Classification by Taxonomist A in 1950 Classification by Taxonomist B in 2000

8. Motivation  Given multiple trees such as described above, how can we provide a simple effective tool to allow taxonomists to effectively and efficiently assert such relationships between taxa?  Presenting the trees side by side as ‘lists’ with a drag ‘n’ drop mechanism for creating relationships will be a simple effective mechanism for the taxonomists to understand / use.

9. Existing techniques  Nesting (e.g. TreeMaps)  Very space efficient.  leaf nodes occlude their parents and ancestors.  Graph layouts  Good at representing structure.  Not space efficient.  Not good at representing taxonomic rank structure.  Adjacency layouts  Space efficient  All nodes are assigned space independent of other nodes.  Shown to be effective for data of this type (e.g. TaxViz).

10. Top-down adjacency layout

11. Concept Relationship Editor  Two taxonomic classifications displayed at a time.  Positioned at either side of the screen.  Oriented so lower ranks face the centre of the screen.  Relationships drawn as curved lines between classifications.  Vertical ‘Focus + Context’ adjacency tree layout….

12. Concept Relationship Editor

13. CRE Tree Layout  Taxa labelled & stacked in ‘lists’ for scanning.  Geometric Distortion  Unfocused taxa are ‘squashed’.  Unfocused taxa remain on-screen so relationships from all focused taxa can be displayed.  Layout algorithm assigns space so that the user can:  Select a closely related node to navigate.  Scan children with a view to adding relationships.  Animation is used to smooth the transition between layouts and prevent the user from becoming disorientated.

14. Layout Algorithm  Nodes can be assigned to one of five groups  F: Focus nodes (i.e. nodes that the user has explicitly selected)  FA: Focus ancestors  FC: Focus children  FS: Focus siblings  FAS: Focus ancestor siblings  They are assigned in this order and cannot be assigned to more than one group.

15. Layout Algorithm  Nodes can be assigned to one of five groups  F: Focus nodes (i.e. nodes that the user has explicitly selected)  FA: Focus ancestors  FC: Focus children  FS: Focus siblings  FAS: Focus ancestor siblings  They are assigned in this order and cannot be assigned to more than one group.

16. Layout Algorithm  Nodes can be assigned to one of five groups  F: Focus nodes (i.e. nodes that the user has explicitly selected)  FA: Focus ancestors  FC: Focus children  FS: Focus siblings  FAS: Focus ancestor siblings  They are assigned in this order and cannot be assigned to more than one group.

17. Layout Algorithm  Assigning vertical space  Priority given to FC then FS then FA nodes.  The sum of all these nodes height is equal to that of the entire display space for the tree.  Higher level nodes (FA) have vertical space that is the sum of their children's space  Nodes below these in the hierarchy have the vertical space from their parent evenly distributed.  Scroll-mode: Total height of FC, FS and FAS is not restricted by screen height. Scrollbar is used to navigate across the hierarchy.

18. Layout Algorithm  Assigning horizontal space  Columns with FC, F and parents of F nodes are prioritised  If the height is above a threshold vertical lettering is used and horizontal spacing is readjusted.  Note: Ellipsis are used when names are too long to be displayed (e.g. CHARADRIIF.. For CHARADRIIFORMES).

19. Navigation

20. Fisheye Lens type effect

21. Asserting relationships

22. User Evaluation  TDWG 2007  Think aloud protocol with 14 representative tasks.  5 users.  42 observations and suggestions for further development.  Ranged from ideas on how the actions should be coordinated to suggested modifications of the terminology used (e.g. ‘Class’ should be used instead of ‘Tax’ for taxonomic classifications).  Also highlighted misunderstandings and things that weren’t so easy for the user to find out how to use (e.g. one user didn’t use the lens scrollbar when concepts were too small and relied on the tool-tip to read concept names).  As users familiarised themselves with the interface they found it easy to navigate the hierarchy of classifications and add relationships between taxa.  Generally positive but highlighted some important usability issues.  Currently planning to undertake a case-study with taxonomists at Cornell Lab of Ornithology working on AOU bird data.

23. Conclusion  We have developed a novel focus + context adjacency layout for hierarchies  Allows multiple groups of nodes to be expanded with their labels readable in horizontally stacked lists.  Can be differentiated from other visualisations in that it allows users to both edit and view different types of relationships between two hierarchies.  During evaluation users found it easy to navigate the hierarchy of classifications and add relationships between taxa.

24. Acknowledgements  NSF funding  SEEK team user requirements  TDWG members user evaluation

25. Navigation

29. Fisheye Lens type effect

30. Asserting relationships