Generalized Formal Models for Faceted User Interfaces Edward Clarkson, Sham Navathe and Jim Foley College of Computing, Georgia Tech Overview 1. Survey of faceted navigation frameworks Observe some common design variances 2. Entity-relationship and relational models for faceted data and queries Generalize to observed design variances Motivation Provide a structured view of faceted navigation design space Conceptual models provide concrete basis for developing new systems in practical terms (RDBMS) Suggest ways for extending state-of-the-art (Is this the right venue?) Background Focus Categorization Categorization Faceted Classification (Ranganathan): Facets Items Categorization Categorization Classification of items into multiple independent (maybe hierarchical) categorizations Faceted Navigation: software UI on FC data “Focus Items” “Facet Values” Faceted Navigation Survey 8 Faceted Browsing systems: Relation Browser, mSpace, Flamenco, Elastic Lists, Humboldt, Parallax, Bungee View and Nested Faceted Browser MS FacetLens (CHI 09) too new Rationale 1. Typical faceted software features (no Tabulator) 2. 3. I.e., filtering via value selection Recency (no Allen, Pollitt) Framework rather than components (no Endeca) Faceted Navigation Survey 1. Visual Design 2. Interaction Design 3. Structural Design Faceted Navigation Survey Visual Design Vertical vs. Horizontal Facet Layout Cardinality Data/Preview Faceted Navigation Survey Interaction Design Selection Cardinality Selection Cascades Faceted Navigation Survey Structural Design Facet Hierarchy Indirect Facets Ind. Facet Facet Ind. Facet Facet Focus Facet Single- vs. MultiFocus Focus Facet Ind. Facet Facet Ind. Facet Ind. Facet Facet Ind. Facet Facet Faceted Navigation Survey Structural Design Facet Hierarchy Indirect Facets Single- vs. MultiFocus Modeling Faceted Metadata Task: can we model the data/queries apparent in faceted navigation software? Take a DB-centric approach: EntityRelationship, relational models Goal: generate a data/query model that generalizes to the variances in our survey nteraction, structural design variance Modeling Faceted Metadata Related Work McGuffin and schraefel (HT 2004) Zhang and Marchionini (JCDL 2004) Dimensional Modeling XFML Modeling Faceted Metadata Background “Some of the discussion on database design is too basic” vs. “Some guidance [reading ER models] would help the layman” attribute Entity relationship Entity Modeling Faceted Metadata Intra-entity ER Data Model Models single entity features (focus or facet): implicit form of entities to follow… Accounts for hierarchical facets, arbitrary item data fields id Name ENTITY Attr1 N child 1 parent Attrn Modeling Faceted Metadata Basic ER Data Model Facet Focus Facet Facet Facet Entities model focus items and facet data Relationships model classification of focus items into facets Faceted Models Extended ER Data Model Accounts for indirect facets Ind. Facet Ind. Facet Facet Straightforward translation to relational models Focus Facet Facet Ind. Facet Facet Modeling Faceted Metadata Relational Query Model Construct relational calculus models that retrieve appropriate Focus items Facet values Account for selection cascades Model Input: facet value selection tuples {(C1, (E1,1,…,E1,k1)),…,(Cn, (En,1,…,En,kn)} Ci = {ci,1,…,ci,ji} and ci,l Ci is the lth selection from FACETi (Ei,1,…,Ei,ki) is the selection path for FACETi Accounts for conjunctive/disjunctive value selections Modeling Faceted Metadata Relational Query Model Iterative development in paper: 1. Queries over basic model 2. Queries with indirect facets Separate queries for focus/facet data 3. Generalized Queries Unified query model for focus/facet data (accounts for for multi-focus environment Accounts for hierarchical selections in focus data Modeling Faceted Metadata Extended Query Model Q : t.Id (e1,1 ) (e1,k1 ) (en ,1 ) (en ,k n ) FACET1 (e1,1 ) FACETk1 (e1,k`1 ) Name the facets that have selections FACET ( e ) FACET ( e ) m k n , 1 m n , k n n (t. parent c p ,1 constraint t. parent c p , j p ) Hierarchical e1,1.Id 1 e1with e1, 2 .Id 1 to e1,k1 1.Id relation 2 e1,k1 .Id 1 t.Id Join facet interest ,1 .Id 2 selection ( e . Id 2 c e . Id 2 c ) 1,Facet k1 1,1 1, k1 1, j1 selection constraint …for each facet selection t.Id en ,1.Id 1 en ,1.Id 2 en , 2 .Id 1 en ,k n 1.Id 2 en ,k n .Id 1 ( e . Id 2 c e . Id 2 c ) n , k1 n ,1 n , k1 n , jn INTEREST (t ) Ind. Facet Ind. Facet Facet Focus Facet Facet Ind. Facet Facet Future Work Extend the model: Expand notion of classifying an item: probabilistic classification? Improve the systems: Performance analysis of interface features Acknowledgements Colleagues Daniel Tunkelang Ian Dickinson/Georgi Kobilarov; Rob Capra/Gary Marchionini; Max Wilson/m.c. schraefel; Marti Hearst Funding Steven Fleming Chair in Telecommunications and DHS/NVAC/SRVAC