2010.05.28 Slide 1 ManyNets Multiple Network Analysis and Visualization Catherine Plaisant Ben...
Preview:
Citation preview
- Slide 1
- Slide 2
- 2010.05.28 Slide 1 ManyNets Multiple Network Analysis and
Visualization Catherine Plaisant Ben Shneiderman Jennifer Golbeck
Manuel Freire-Moran manuel.freire@uam.es 2010.05.18 Awalin Nabila
Miguel Rios Manuel Freire
- Slide 3
- 2010.05.28 Slide 2 1 social network What about comparing
thousands?
- Slide 4
- 2010.05.28 Slide 3 1 row = 1 network Columns = network features
(metrics, distributions) Column summaries = interactive overviews
ManyNets SocialAction [Perer08]
- Slide 5
- 2010.05.28 Slide 4 1 row = 1 network Columns = network features
(metrics, distributions) Column summaries = interactive overviews
ManyNets SocialAction [Perer08]
- Slide 6
- 2010.05.28 Slide 5 1 row = 1 network Columns = network features
(metrics, distributions) Column summaries = interactive overviews
ManyNets SocialAction [Perer08]
- Slide 7
- 2010.05.28 Slide 6 1 row = 1 network Columns = network features
(metrics, distributions) Column summaries = interactive overviews
ManyNets SocialAction [Perer08]
- Slide 8
- 2010.05.28 Slide 7 Split large networks to compare parts
Multiple criteria sort, Filter using custom expressions Tight
coupling with node-link diagrams e.g. all ego-networks FilmTrust
[Golbeck06]
- Slide 9
- 2010.05.28 Slide 8
- Slide 10
- 2010.05.28 Slide 9 row selection & column overviews
- Slide 11
- 2010.05.28 Slide 10
- Slide 12
- 2010.05.28 Slide 11 1 row = 1 network Columns = network
features (metrics, distributions) Column summaries = interactive
overviews Target users: network analysts ManyNets
- Slide 13
- 2010.05.28 Slide 12 Motivation Analysis of separate networks:
compare a set of networks Analysis of parts of a single network:
divide and conquer Local neighborhoods (ego networks) within a
social network Compare larger neighborhoods (clusters or
communities) Find prevalence of certain network motifs Compare
sub-networks with certain attributes (eg.: time-slices) Analysis of
multi-modal networks Handle networks with multiple types of nodes
and edges Generate new edges (two users are connected if)
- Slide 14
- 2010.05.28 Slide 13 separate networks example Facebook networks
from 5 US universities, from [Traud09]
- Slide 15
- 2010.05.28 Slide 14 separate networks example Facebook networks
from 5 US universities, from [Traud09]
- Slide 16
- 2010.05.28 Slide 15 separate networks example Facebook networks
from 5 US universities, from [Traud09]
- Slide 17
- 2010.05.28 Slide 16 separate networks example
- Slide 18
- 2010.05.28 Slide 17 separate networks example
- Slide 19
- 2010.05.28 Slide 18 Motivation Analysis of separate networks:
compare a set of networks Analysis of parts of a single network:
divide and conquer Local neighborhoods (ego networks) within a
social network Compare larger neighborhoods (clusters or
communities) Find prevalence of certain network motifs Compare
sub-networks with certain attributes (e.g.: time-slices) Analysis
of multi-modal networks Handle networks with multiple types of
nodes and edges Generate new edges (two users are connected
if)
- Slide 20
- 2010.05.28 Slide 19 single network example Nodes are users
Links are trust ratings in other users film rating expertise
JoeMary 10 PeterPaul 8 2 Mark Tim 8 9 Ed ? FilmTrust
[Golbeck06]
- Slide 21
- 2010.05.28 Slide 20 Mark ego network radius 0
- Slide 22
- 2010.05.28 Slide 21 ego network radius 1 Mark TimEd Joe Mary
Peter Paul
- Slide 23
- 2010.05.28 Slide 22 ego network radius 1.5 Mark TimEd Joe Mary
Peter Paul
- Slide 24
- 2010.05.28 Slide 23 ego network radius 2 Mark TimEd Joe Mary
Peter Paul Liz Ben Jane Beth Tom
- Slide 25
- 2010.05.28 Slide 24 Q: are big ego nets similar to small ones?
picture of trust distribution in big ego nets (large neighborhood)
picture of trust distribution in small ego nets (small
neighborhood)
- Slide 26
- 2010.05.28 Slide 25
- Slide 27
- 2010.05.28 Slide 26 are big ego nets similar to small ones?
picture of trust distribution in big ego nets (large neighborhood)
picture of trust distribution in small ego nets (small
neighborhood)
- Slide 28
- 2010.05.28 Slide 27 Motivation Analysis of separate networks:
compare a set of networks Analysis of parts of a single network:
divide and conquer Local neighborhoods (ego networks) within a
social network Compare larger neighborhoods (clusters or
communities) Find prevalence of certain network motifs Compare
sub-networks with certain attributes (eg.: time-slices) Analysis of
multi-modal networks Handle networks with multiple types of nodes
and edges Generate new edges (two users are connected if)
- Slide 29
- 2010.05.28 Slide 28 multi-modal network example BobAlice Jaws
Star-wars trust = 8/10 rating = 4/5 rating = 3/5 rating = 2/5
- Slide 30
- 2010.05.28 Slide 29 Interface Support for multi-modal networks
Schemas Table levels Columns (network metrics, features) can be
removed, rearranged, added From menu Via user-specified expression
Filter and sort Details on demand in side-pane, tooltips Create new
relationships, access the overall schema
- Slide 31
- 2010.05.28 Slide 30 schemas BobAlice Jaws Star-wars trust =
8/10 rating = 4/5 rating = 3/5 rating = 2/5 user film trust rating
FilmTrust Schema
- Slide 32
- 2010.05.28 Slide 31 user film trust rating
- Slide 33
- 2010.05.28 Slide 32 user film trust rating
- Slide 34
- 2010.05.28 Slide 33 user film trust rating
- Slide 35
- 2010.05.28 Slide 34 user film trust rating
- Slide 36
- 2010.05.28 Slide 35 multiple node and edge types: levels Lowest
level: entity and relationship tables Entities are stand-alone, can
be used as nodes Relationships relate two entities, map to edges
Inside a network: node and edge tables Nodes come from entities
Edges come from relationships Can mix multiple entities,
relationships in a network: multi-relational or multi-modal Network
tables Each row is a network
- Slide 37
- 2010.05.28 Slide 36 Interface Support for multi-modal networks
Schemas Table levels Columns (network metrics, features) can be
removed, rearranged, added From menu Via user-specified expression
Filter and sort table Details on demand in side-pane, tooltips
Create new relationships, access the overall schema
- Slide 38
- 2010.05.28 Slide 37
- Slide 39
- 2010.05.28 Slide 38
- Slide 40
- 2010.05.28 Slide 39 Interface Support for multi-modal networks
Schemas Table levels Columns (network metrics, features) can be
removed, rearranged, added From menu Via user-specified expression
Filter and sort Details on demand in side-pane, tooltips Advanced
column overviews Create new relationships, access the overall
schema
- Slide 41
- 2010.05.28 Slide 40 Overviews of Distribution Columns ManyNets
Overviews [Sopan10 / under review]
- Slide 42
- 2010.05.28 Slide 41 Overviews of Distribution Columns ManyNets
Overviews [Sopan10 / under review]
- Slide 43
- 2010.05.28 Slide 42 Interface Support for multi-modal networks
Schemas Table levels Columns (network metrics, features) can be
removed, rearranged, added From menu Via user-specified expression
Filter and sort table Details on demand in side-pane, tooltips
Create new relationships, access the schema
- Slide 44
- 2010.05.28 Slide 43 Deriving new relationships BobAlice Jaws
Star-wars trust = 8/10 rating = 4/5 rating = 3/5 rating = 2/5 user
film trust rating FilmTrust Schema
- Slide 45
- 2010.05.28 Slide 44 Deriving new relationships BobAlice Jaws
Star-wars trust = 8/10 rating = 4/5 rating = 3/5 rating = 2/5 user
film trust rating Extended Schema Co-rated weight = 1
- Slide 46
- 2010.05.28 Slide 45 Deriving new relationships Build new
relationships on the fly Extend schema with each relationship
Retain access to original data Compare resulting networks to each
other user film trust rating Co-rated Good predictor for
- Slide 47
- 2010.05.28 Slide 46 Validation Original ManyNets (presented at
CHI 2010) Case Study on FilmTrust with domain expert Formative
usability test (7 users) ManyNets2 (work in progress) NSF grant
data Your dataset here!
- Slide 48
- 2010.05.28 Slide 47 Conclussion Multimodal network analysis is
hard ManyNets can help! build and explore sets of networks split,
filter, rank, overview, drill, elide, synthesize Reveals patterns
within network attributes Does so interactively, allowing
exploratory search Development page (application, datasets, manual)
tangow.ii.uam.es/mn/ open-source, feedback welcome! (please contact
us) Academic page (publications, demo videos)
www.cs.umd.edu/projects/hcil/manynets/ Acknowledgements Partial
support from Lockheed Martin Manuel Freire supported by Fulbright
Scholarship