Goal-based Recommendation utilizing Latent Dirichlet Allocation

Introduction
Methodology
Results & Findings
Goal-based messages Recommendation
utilizing Latent Dirichlet Allocation
Sébastien Louvigné, Yoshihiro Kato, Neil Rubens, and
Maomi Ueno
Graduate School of Information Systems
The University of Electro-Communications
Tokyo, Japan
Jul 8, 2014
Sébastien Louvigné (louvigne@ai.is.uec.ac.jp) - UEC Tokyo IEEE - ICALT 2014. Athens, Greece.

Introduction
Methodology
Results & Findings
Outline
1 Introduction
Learning Goal & Purpose
Problem Statement
Proposed research
2 Methodology
Goal-based Data
Goal & Purpose Recommendation
Implementation
3 Results & Findings
LDA results
Perplexity
Learners Evaluation
Discussion

Introduction
Methodology
Results & Findings
Learning Goal & Purpose
Problem Statement
Proposed research
Goals for Learning
Goal enhances Learning
Providing a sense of direction to attain specic standards.
Critical motivator (personal emotions, beliefs).
(Schunk et al. 2002)
Goal orientations
Refer to purposes for engaging in achievement tasks.
(Pintrich, 2003)

Introduction
Methodology
Results Findings
Learning Goal Purpose
Problem Statement
Proposed research
Goal Purpose
Denitions
1 Goal: terminal point towards which action is directed (e.g.
master a language).
2 Purpose: provides the psychological force to attain a goal
(i.e. reasons for learning).
Goals ecient when linked with learner's needs (purpose for
learning).
Learners have dierent purposes (conceptual perceptions).
Goal orientations have dierent eects on intrinsic
motivation.

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Problem Statement
Why learning?
Highly structured education Syllabus states objectives.
Learners have their own conceptions Often unrelated with
formal education.
Goal Orientation should be set properly
Risk of conict / discouragement / harm intrinsic motivation.

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Goal Theory
Goal Setting
Goal properties inuencing learning performance and intrinsic
motivation (Locke Latham, 1990; Zimmerman et al. 1992; Bekele, 2010).
Self-set goals often motivate better than assigned goals.
Goal orientations
Mastery goals (internal) vs. Performance goals (external)
(Ames, 1992).
Task involvement vs. Ego involvement (Nicholls, 1979).
Approach / Avoidance distinction (Elliot, 1997).
Personal goals
Focusing on Personal conceptions (Self theories).

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Proposed approach
This research: Using Social Context for motivation
Sharing goal orientation (goal content + purpose) with others.
Adopting new purposes for learning.

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Research purpose
Research Question
How to use Social Networks (i.e. peers) to improve
learning motivation?
Hypothesis
Learners enhance motivation by observing goal purposes from
other peers.
Diversity of goal purposes aects learners' motivation.

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Proposed System
Recommendation System
Diversity of Learning Purposes from peers.

Introduction
Methodology
Results Findings
Problem Statement
Proposed research
Learning in Social Settings
Previous works
Social Cognitive Theory
Knowledge acquisition by observing others (Bandura, 1988).
Social constructivism
Groups of learners construct knowledge collaboratively
(Vygotsky, 1978).
Cognitive apprenticeship
Modelling, Scaolding, Reecting knowledge (Collins, 2006).
This research: enhance motivation
Using peers' motivational contents to enhance purpose for
learning.

Introduction
Methodology
Results Findings
Goal-based Data
Goal Purpose Recommendation
Implementation
Largescale Dataset
Social Media: Twitter
Short text messages
Metadata (e.g. user prole, social network)
Large amount of data publicly available
(Louvigné et al. 2012; Shi Louvigné, 2014)

Introduction
Methodology
Results Findings
Goal-based Data
Implementation
Latent Dirichlet Allocation (LDA)
Probabilistic model for collections of discrete data (Blei et al. 2003)
α : Dirichlet parameter prior on
per-document topic distribution
β : Dirichlet parameter prior on
per-topic word distribution
(Asuncion et al. 2009)
Documents: Mixture of topics
Full conditional: P(zi = j|z−i ,w) ∝
n
(w
i
)
−i ,j
+β
n
(.)
j
+W β
(n(di )
−i,j
+α)
Dirichlet: ˆφ
(w )
j
=
n
(w )
j
+β
n
(.)
j
+W β
(Griths Steyvers. 2004)

Introduction
Methodology
Results Findings
Goal-based Data
Implementation
Data Organization
How to use the database (goal + purpose)
LDA: Find various goals purposes within a same learning
subject.
Probability distribution: message belonging to a topic.

Introduction
Methodology
Results Findings
Goal-based Data
Implementation
Goal-based Recommendation
Process
Recommending Learning Purpose messages based on:
Similarity: similar goal.
Diversity: various purposes.

Introduction
Methodology
Results Findings
Goal-based Data
Implementation
User Interface

Introduction
Methodology
Results Findings
Goal-based Data
Implementation
Goal Prole

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
LDA results
Finding various topics
Diverse topics within dataset of goal-based Twitter messages

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Perplexity
Finding optimal number of topics
Dierent optimal number of topics for each learning subject.
Not related with number of messages.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Goal attributes for Motivation Evaluation
Goal-Setting: Attributes inuencing learning and
performance (Locke, 1990; Zimmerman et al. 1992; Bekele, 2010).
Goal attributes
Leading eventually to personal satisfaction (Fulllment).
Fulllment and achievement motivation: important
success factors in learning.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Goal attributes evaluation
Before peers' messages observation
High perception: Importance, attainability, diculty.
Low perception: Commitment, performance, fulllment,
condence.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Goal attributes evaluation
After observation
Similarity: slight increase in learner's perception of goal
attributes.
Diversity: higher increase in specicity and commitment,
decrease for diculty.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Conclusion
Using Social Context to enhance learning motivation
1 Observing goal purposes from peers.
Adopt new purposes.
2 Diversity of goal purposes.
Aect intrinsic motivation.
Results
LDA for learning purposes recommendation
Various topics (i.e. purposes) for a same learning subject (i.e.
mastery goal).
Observing goal purposes from peers
Similarity: conrms learner perception on goal,
Diversity: improve specicity, commitment.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Future works
LDA
e.g. Short text analysis, Including grammatical features
Motivation evaluation
Long term experiment
Free choice
Evaluation from peers

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Bibliography
E. A. Locke (1996), Motivation through conscious goal setting. Applied
Preventive Psychology.
D. H. Schunk, J. L. Meece, and P. R. Pintrich (2002), Goals and Goal
Orientations. Motivation in Education: Theory, Research, and Applications.
P. R. Pintrich (2003), A Motivational Science Perspective on the Role of
Student Motivation in Learning and Teaching Contexts. Journal of Educational
Psychology.
E. A. Locke, and G. P. Latham (2002), Building a practically useful theory of
goal setting and task motivation: A 35-year odyssey. American Psychologist.
D. M. Blei, A. Y. Ng, and M. I. Jordan (2003), Latent Dirichlet Allocation.
Journal of Machine Learning Research.
T. L. Griths, and M. Steyvers (2004), Finding scientic topics. National
academy of Sciences of the United States of America.
S. Louvigné, N. Rubens, F. Anma, and T. Okamoto (2012), Utilizing Social
Media for Goal Setting based on Observational Learning. IEEE Icalt 2012.
J. Shi, and S. Louvigné (2014), Goal-Setting and Meaning-Making in Mined
Dataset of Tweets Using SFG Approach. Journal of Electrical Engineering.

Introduction
Methodology
Results Findings
LDA results
Perplexity
Learners Evaluation
Discussion
Thank you

Goal-based Recommendation utilizing Latent Dirichlet Allocation

More Related Content

What's hot (20)

Viewers also liked (17)

Similar to Goal-based Recommendation utilizing Latent Dirichlet Allocation (20)

Recently uploaded (20)

Goal-based Recommendation utilizing Latent Dirichlet Allocation