An Overview of Recent Developments in Intelligent e-Textbooks and Reading Analytics
1. An Overview
of Recent
Developments
in Intelligent
e-Textbooks
and Reading
Analytics
BY DAVID BOULANGER
AND
VIVEKANANDAN KUMAR
Athabasca University | Canada
3. DATASETS
3
# of events Events /
reader
# of
readers
Reading
episode Domain Software Type of reading
2,812,727 939 2993 Semester (43 courses) BookLooper Lecture
567,193 7988 71 Semester
Human Computer
Interaction,
Information Retrieval
Reading Circle
Textbook,
research
publication, etc.
129,451 555 233 Semester (11 courses) CourseSmart Textbook
75,748 276 274 Semester Social Sciences,
Business, Education - Textbook
65,755 608 108 Semester (2 courses) BookRoll Lecture
10,994 - - ~1 year (110 different magazine
issues)
Viewerplus + APP-
BI Magazine
10,188 35 289 Semester Interactive Systems
Design
AnnotatED+,
Reading Circle Textbook
~7200 109 66 Semester Research Methods - Textbook
1370 80 17 1.5 hours Educational Technology DITeL Journal article
65 7 9 30 minutes Introductory Biology - Textbook
In-house readers except for CourseSmart
4 types of reading
The semester is the typical period during
which reading behavior is evaluated.
4. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
4
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Feature
SelectionSmart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
DOCUMENT
• Page
• Section
• Chapter
Raw Data
Interaction
Data
Self-Reported
Data
DOCUMENT FORMAT
• PDF
• EPUB
• Word documents
• Web documents
• Any reading resource
5. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
5
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Feature
SelectionSmart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
Raw Data
Interaction
Data
Self-Reported
Data
HARDWARE
• Augmented reality
• Eye tracker
• Mobile
• Remote
• Smartphones
• Tablets
• E-readers
• Computers
SOFTWARE
• BookLooper
• Reading Circle
• CourseSmart
• BookRoll
• Viewerplus
• AnnotatED+
• DITeL
• Adobe Reader
• Kindle
6. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
6
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
INTERACTION DATA
• Next page
• Previous page
• Jump to page
• Annotations
• Marking of
unknown words
• Comments
• Highlighting
• Underlining
• Changing marker
color
• Zooming in/out
• Searching
• Search jump
• Screen orientation
• Portrait
• Landscape
• Scrolling
• Bookmarks
• Opening, exiting,
minimizing e-textbooks
SELF-REPORTED DATA
• Survey
• Interview
• Questionnaires
Raw Data
Interaction
Data
Self-Reported
Data
• Online
• Offline
Feature
Selection
7. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
7
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
METRICS
• # of blinks
• Distances of eye
movements
• Coordinates of eye gazes
• Fixations
• Saccades
• Time of adoption
• Reading speed
• Engagement level
• Reading session
• Attention
• Reading/visit time
• Last pages read
• Pages previewedRaw Data
Interaction
Data
Self-Reported
Data
Feature
Selection
8. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
8
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
READING BEHAVIOR
• Progressive sequential
analysis
• Self-reported behavior
• Perception gap
Raw Data
Interaction
Data
Self-Reported
Data
ENGAGEMENT MODEL
• Blocked linear regression
• Clustering
• None
• Low
• Medium
• High
GAMING BEHAVIOR
• Classifier (gaming,
normal)
• Logistic regression
• KNN
• Naïve Bayes
• Decision tree
• SVM
Feature
Selection
9. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
9
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
TECHNIQUES
• Gradient tree boosting
• Survival analysis
• PCA
• Regression
• Simple linear
• Non-linear
• Mixed-model linear
• Deep learning
• Stochastic block model
• Descriptive statistics
• Classification
• Random forest
• KNN
• Support vector machine
Raw Data
Interaction
Data
Self-Reported
Data FEATURE SELECTION
• Random forest regressor
• F-regression
• K-means transformations
• t-test
ACCURACY
• Training/testing set
• Cross-validation
• AUC
• RMSE/MSE
Feature
Selection
10. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
10
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
SMART FEATURES
• Prediction of student
performance
• Identification of at-risk
students/drop-outs
• Real-time improvement
of learning materials
• Provision of teacher
annotations
• Measurement of
reader’s interest and
competence
• Assessment of a
concept’s difficulty level
• Automatic correction of
students’ answers
• Provision of formative
feedback
• Recommendation of
next topics to be learned
• Recommendation of
effective learning
strategies
• Real-time lecture
supporting system
• Many more …
Raw Data
Interaction
Data
Self-Reported
Data
Feature
Selection
11. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
11
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
INTERACTIVE COMPONENTS
• Multimedia/visuals
• Video
• Sketch
• Animation
• Diagram
• Problems to solve
• Collaborative
• Q&A
• Quizzes (multiple-
choice/open-ended
questions)
• Explanations of key
concepts
• Examples with different
parameters
Raw Data
Interaction
Data
Self-Reported
Data
Feature
Selection
12. CONCEPT MAP OF READING ANALYTICS & INTELLIGENT E-TEXTBOOKS
12
E-Reader
Document
Document
Format
Hardware
Reading
Behavior
Software
Metrics
Engagement
Model
Gaming
BehaviorTechniques
Accuracy
Smart
Features
Interactive
Components
Adaptive
Engine
Instructional
Modeling
Student
Modeling
Content
Modeling
Student
ADAPTIVE ENGINE
• Student model
• Learning style
• Instructional model
• Knowledge level
• Content model
• Curriculum
Raw Data
Interaction
Data
Self-Reported
Data
STUDENT MODELING
• Transfer GPA
• Demographics
CONTENT MODELING
• Concept-based
hyperspace organization
• Automatic
• Bag-of-words
• LDA
• Manual
INSTRUCTIONAL MODELING
• Knowledge Tracing
• Hidden Markov model
• Behavior model
• Performance model
• Behavior-performance
model
Feature
Selection
13. FUTURE WORK
13
▪ Benchmark datasets on both reading activities and heterogeneous learning
activities
▪ Further experimentation of deep learning techniques (e.g., multimodal RNNs)
▪ Analysis of the reading process in the broader frame of the learning process
(e.g., how reading is used as a learning strategy)
▪ Impact of e-textbook’s smart features on reading and learning performance
▪ Discovery of optimal reading behaviors
▪ E-textbook’s interactive components implemented in virtual/augmented reality
environments
▪ Intelligent printed textbooks/documents through augmented reality
▪ Intelligent adaptivity to nurture self-regulatory traits
▪ Adaptivity (computer-driven) vs. adaptability (human-driven)
14. http://www.bbc.com/capital/story/20190127-humanics-a-way-to-robot-proof-your-career
LEARNING ANALYTICS, COGNIFICATION & HUMANICS
14
Measuring the impact of
learning analytics on
learning performance
• Accelerating learning time
• Increasing knowledge retention
• Formation of soft skills
• Greater transfer of learning Humanics-oriented life-long learning
Ensuring humans remain relevant
in the world of AI.
• Mastery of learning content:
• Traditional literacies
• Tech literacy
• Data literacy
• Human literacy
• Nurture and assessment of cognitive
capacities:
• Creativity
• Mental flexibility
• Critical thinking
• Systems thinking
• Leadership
Cognification of learning
The process of making learning
objects and environments
increasingly and ethically
smarter.
• Intelligent textbook
• Competence assessment
• Automated scoring
• Formative feedback
• Remedial interventions
“A generation ago, the half-life of a skill
was about 26 years. Today, it’s 4 and
half years and dropping.” – Indranil Roy
15. Learner Instructor
Organization Governance
Stakeholders
IoT Conversations
Data
Artisanship
Dashboard
Insights
Descriptive
Diagnostic
Predictive
Prescriptive
Dimensions
Causality
Observational
Randomized
Longitudinal
Meta-analysis
Meta-analysis
Meta-analysis
Data Mining &
Research Methods
Knowledge
Base
“An ethics-bound, semi-autonomous,
and trust-enabled human-AI fusion that
measures and advances knowledge
boundaries in human learning.”
A DEFINITION OF LEARNING ANALYTICS
15
Kumar, V., Boulanger, D., Fraser, S.: Inferring Causal Effects from Learning Analytics: Discovering the Nature of Bias (2019).
16. Instructional Model
Topic E
Topic D
Topic G
Topic H
Topic J
Assessment of Decision Making
Content Model
Expert Machine Learning
Topics ready
to be learned
Topic D
Topic E
Learner Model
Adaptive Engine
Learning Object
Repository
Trust Model
Talent
SRL
Learning
Style
Grit
Interest
Engage-
ment
Demo-
graphics
Learning
Strategy
Moti-
vation
Prior
Know-
ledge
Learner/Teacher/AI Decision Making
Pedagogical Strategy
2x
E
1x
D
3x
G
5x
H
2x
J
ADAPTIVITY
16