Multi-model Adaptive Spatial Hypermedia Luis Francisco-Revilla Department of Computer Science Texas A&M University September 22, 2004
1. Agenda Problem Related Work Theoretical contributions Conceptual Architecture Content and Adaptation Conflict Management Practical contributions WARP Evaluation Results Conclusions Future Work
Problem Spatial Hypermedia Adaptive Hypermedia Fixed documents result in rigid presentations of the information
Problem Hypermedia Multi-model Adaptive Hypermedia Adaptive Hypermedia Personalize presentations Adapt presentation to multiple aspects
Problem Spatial Hypermedia Hypermedia Increase expressiveness of the medium Can represent relationships between documents with varying degree of formality and ambiguity
Problem Spatial Hypermedia M ulti-model A daptive S patial H ypermedia Multi-model Adaptive Hypermedia Adaptive Hypermedia
Related Work Spatial hypermedia Adaptive hypermedia Multi-model adaptive hypermedia Multi-agent systems Visual design
Theoretical Contributions: Architecture Hypermedia Platform Parser Analyzer Transformer Composites Metrics Transformations Behaviors M1M1 M2M2 MnMn Hypermedia Generator Contents Dynamic Adaptive Generative Interactive Models
Space, Objects and Relationships Object Atomic Document Composite Relationship Quality Association Scope Implicit Relative Intra-space Explicit Absolute Extra-space Spatial Hypermedia vs. Navigational Hypermedia Space Nature Topology Instantiation Dimensions Connectivity Co-location Rendering
Adaptation Ontology Semantic Relational Spatial Meta-adaptation
Conflict Management When multiple adaptations cannot be simultaneously represented Avoidance Detection Resolution Managing conflicts is more than solving conflicts Object Model 1: highlight Model 2: hide
Conflict Avoidance Augment medium expressiveness Dynamically map high-level methods low-level techniques Embrace ambiguity
Conflict Detection Conflict propagation Scope of conflicts Spatial parser Types of conflicts Object Sibling Parent Parent-Child Parent-Parent AB 2 1 Suggestion 1 Suggestion
Conflict Resolution (1) Merge suggestions Multiple strategies Weighted average Suggestion strength Suggestion confidence Model confidence Heuristic best Object Model 1: emphasize Model 2: de-emphasize Object average suggestions
Determine mapping from adaptation methods to techniques Balancing author and reader control Specify mapping and resolution strategies Conflict Resolution (2) emphasize Object emphasize Object emphasize
Practical Contributions: WARP Multi-model Adaptive Spatial Hypermedia Executes in a Web-browser Novel features Personal readings Annotations Transclusion links Behaviors
Reading and interacting Interaction and personalization of spatial hypertexts (moving, modifying, etc.) Readers can annotate documents Save and restore users’ reading state Reading = Interacting
Transclusion Links Web pages WARP spaces Import and export
Behaviors User actions and system adaptations can affect existing spatial structures Spatial parser identifies structures Behaviors can preserve spatial relationships
Multiple Independent Models Complexity and scalability Easier knowledge engineering Portability and reutilization Amortization of costs Privacy and distribution Control over personal models Not Very So-so Not So-so Very NotSo-soVery NotSo-soVery NotSo-soVery NotSo-soVery
Adaptation Process (1) Objects prior to adaptation Platform Parser Analyzer Transformer
Adaptation Process (2) Inference of implicit structures Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models
Adaptation Process (3) Context inference and conflict prevention Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models
Adaptation Process (4) Suggestion of adaptations Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models
Adaptation Process (5) Transformation and adaptations Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models
Adaptation Process (6) Extended conflict detection Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models Conflict?
Adaptation Process (7) Alternative creation Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models Alternatives
Adaptation Process (8) Evaluation of alternatives Platform Parser Analyzer Transformer M1M1 M2M2 MnMn Models OK?
Adaptation Process (9) Final adaptation Platform Parser Analyzer Transformer
Evaluation Comparative study Non-adaptive spatial hypermedia Adaptive spatial hypermedia Task model (Web page requirements) User model (HTML knowledge) Investigate the effects of adaptation in the process of reading spatial hypermedia Usability
Evaluation 16 participants (8 Novice, 8 Expert) Author a Web page Spatial hypertext about HTML / XHTML Text editor 90 minute limit User scenario and motivation TimeActivity 15 minutes Training in software tools (WARP and authoring environment) 5 minutes Completing the computer and Web expertise questionnaire 20 minutes Completing the HTML and XHTML questionnaire 90 minutesAuthoring Web page 10 minutes Completing the questionnaire about use of the system 10 minutesInterview 2:30 hours
Document design Content Kennedy and Musciano “HTML & XHTML: The Definitive Guide” O’Reilley’s, 2002 Layout Encapsulate topics and subtopics Visually reflect the structure of the information Limited dynamic behaviors Adaptive behaviors Multiple visual cues – size, font-size, glow, alpha-blur, zooming
Initial Interface (not adapted)
Initial interface (adapted)
Results Qualitative Emergent reading strategies Observations, interviews, comments Quantitative Significant differences Adaptive vs. non-adaptive Novice vs. advanced Assessments of the quality of the Web pages
Layout Easy navigation of large information spaces Effective concept encapsulation Reflect information structure “I really like that I can see all of the chapters”
Moving and rearranging …to indicate what is being read or what has been read..to indicate “what is more important” …in order to “see both and compare” …“for reference”
Collections Maximizing sub-collections to bookmark sections Minimizing collections as “I’m done with that!”
Informed link traversals Browsing before committing to maximizing collections “You are not clicking on a bunch of links that may or may not have what you are looking for”
Adaptation Adaptation changed the way people read Implementation guidelines: Using multiple visual cues Allow readers to maintain control of the process “What red glow?”
Quantitative Results Evaluation of Web pages Content Presentation Overall Significant differences Expertise level Adaptation Predefined metrics Overallp = Contentp = Presentationp = Adaptive vs. Non-adaptive Overallp = Contentp = Presentationp = Novice vs. Advanced Overallp = Contentp = Presentationp = Expertise-Adaptation Interaction
Expertise and Overall scores Some clustering Correlation lower than expected Expertise = HTML knowledge * (1 + Previous knowledge) 0.684Adaptive Spatial Hypermedia 0.894Non-adaptive Spatial Hypermedia Correlation between expertise and overall score
Conclusions: Theoretical contributions Multi-model Adaptive Spatial Hypermedia framework Conflict management Enables the use of multiple independent models Prevention Detection Resolution
Conclusions Demonstrates the benefits of: Adaptive spatial hypertext Adaptation based on multiple independent models Adaptive spatial hypermedia is an effective medium that facilitates the navigation of large information spaces Adaptation affects the process of reading in spatial hypermedia
Conclusions: Practical contributions WARP: Full-fledged implementation of Multi-model Adaptive Spatial Hypertext Web based Personal readings Annotations Transclusion links Behaviors
Future Work Cool interface stuff Mixed media, behaviors, etc. Improve support for interaction and authoring CSCW implications Front-end component for systems Interface for digital libraries and search engines Software engineering Meta-adaptation Automatic adjustment of adaptation parameters
Questions?