An Introduction to TuTalk: Developing Dialogue Agents for Learning Applications

1
An Introduction to TuTalk Developing Dialogue
Agents for Learning Applications

• Pamela Jordan
• University of Pittsburgh
• Learning Research and Development Center

2
Dialogues in Learning Applications

• Language practice
• Peer collaborations
• Short answer tutorial dialogue
• AutoTutor
• Circsim-Tutor
• TuTalk

3
AutoTutor - Physics
4
Circsim-Tutor
5
Cordillera (built using TuTalk)
6
ProPL (built using TuTalk)
7
What is TuTalk?

• Dialogue system construction tools for learning
  applications
• Authoring tools for creating, testing and
  maintaining an artificial dialogue partner/agent
• Dialogue system server that
• Conducts dialogues with multiple students
• Provides communication protocol for integrating
  with interfaces and applications (Several how-to
  examples provided)
• Provides experiment management tools

8
Finite state dialogue management
gravitational,
ltanything elsegt
runners
9
Example dialogue
Subdialogue
10
TuTalk Dialogue System Usage History
Experiment or System Name Domain Number of experiments
Andes-Atlas Quantitative physics 3
Why2-Atlas Qualitative physics 3
ITSpoke Qualitative physics 4
Reflective follow-up Quantitative physics 2
ProPL Computer Science 1
CycleTalk Thermodynamics 2
Collaborative problem solving Math (6th grade) 1
Cordillera Quantitative qualitative physics 2
18
11
What do you have to do to create a TuTalk
dialogue agent?

• Write domain content in form of natural language
  dialogue turns (e.g. elicit or tell)
• Write an ideal dialogue on a topic
• Write expected short answer student responses
  (correct, not correct)
• Write subdialogues for expected student responses
  that are
• Partially correct/incomplete
• Partially incorrect
• Overly vague
• Overly specific
• Correct but premature

12
Authoring, Previewing and Testing Demo
13
Authoring, Previewing and Testing Demo
14
When are short answer dialogues
appropriate/inappropriate?

• Appropriate for
• practicing some dialogue skills
• conceptual discussions
• scaffolding problem solving
• identifying addressing gaps in student
  understanding only as needed (hints, examples)
• Not appropriate for
• assessing deep understanding
• addressing grammar problems in language
• content delivery printed text is more efficient
• student-only initiative (use CTAT instead)

15
Past Summer School TuTalk projects

• Language tutoring
• Coaching military trainees to follow one required
  communications protocol
• Giving ESL learners dialogue practice
• Coaching student is proper use of two Chinese
  lexical items that depend on context

16
PSLC Summer School 2007

• Using TuTalk to build a tutor for
• Chinese pronunciation
• Wenyan Zhou, Vanderbilt University
• Tiffany Taylor, George Mason University

17
Implementation

• Choosing the wrong pinyin provides feedback and
  choosing the right pinyin but the wrong tone
  leads to remediation

18
Past Summer School TuTalk projects

• Conceptual tutoring
• Coaching elementary school students in
  qualitative reasoning skills
• Coaching students on loop constructs in
  programming
• Coaching students in the solution of monomials
• Coaching students on Pythagorean Theorem

19
CT Percent Tutor Metacognitive TuTalk Prompts

• Primary school aged students
• Learning Objectives
• Learning fractions, percentages, and ratios
• Translating word problem into an equation
• ITS Roles
• CT
• Model Tracing
• Facilitates problem steps
• Detect a suboptimal two buggy paths
• TuTalk
• Metacognitive prompts
• Encourages self-monitoring goal settings
• Facilitates analogous solution strategies

Dr. Baba Kofi Weusijanae-mail
babaw_at_u.washington.edu Yvette Aquie-mail
yvette.aqui_at_unlv.edu
20
Demo CTAT Incorrect Path 1Percent Conversion
Incorrect 2
21
Demo TuTalk Script 1 , Example 1
22
Tuesday TuTalk Track

• Track Lecture Basics of Authoring TuTalk
  Dialogue Agents
• Review and expand on basic authoring with GUI
• Introduce alternative scripting language for
  authoring
• Hands-on Create a simple TuTalk Dialogue Agent
• Do section 3.3 of TuTalk Authoring Interface
  Users Guide (can do sections 3.1 and 3.2 first
  if you prefer)
• Project Locate relevant dialogues or collect
  small sample of dialogues (available corpora
  http//andes3.lrdc.pitt.edu/TuTalk/corpora/)
• Track Lecture The Methodology of Authoring
  Dialogues
• Dialogue authoring methodologies  
• Advice/findings on effective learning dialogues
• Project Identify problem solving goals to cover
  in dialogue
• Project Begin dialogue authoring (mainlines of
  reasoning w/ correct and default follow-up)
• Group Lectures Think alouds difficulty factors
  assessment, Educational Data Mining

23
Wednesday TuTalk Track

• Track Lecture Advanced TuTalk Dialogue Agents
• Discuss TuTalk server and its support tools
• Explore additional authoring features (e.g.
  looping, optional steps)
• Project Test and refine dialogue goals
  implemented dialogues
• Project Add more expected student responses
  follow-ups
• Project Test and refine additions
• Project Create alternative ways of achieving
  dialogue goals subgoals, e.g.
• Version of dialogue for advanced student (e.g.
  agent does or summarizes easy steps and scaffolds
  harder steps, ask for justifications)
• Version for less advanced student (e.g. agent
  scaffolds easy steps and models the harder steps,
  agent explains justifications for steps)
• Group Lectures Issues in transfer learning,
  Cognitive principles in tutor design
• Group Demos (in parallel) ML for building a
  cognitive tutor, ESL demo

24
Thursday TuTalk Track

• Project Test latest dialogues and refine
• Project Finalize dialogue agent
• Final testing
• Set up an interface to demo agent for poster
  session
• Integrate with other tools that are triggers for
  dialogue goals
• Project Prepare posters, presentations
• Lecture Socio-cultural perspectives on learning

25
TuTalk Development Team

• Authoring tools
• Carolyn Rosé
• Yue Cui (Jenny)
• Rohit Kumar
• Dialogue system server
• Pam Jordan
• Brian Hall (Moses)
• Michael Ringenberg

26
TuTalk Summer School Team

• Pam Jordan (dialogue system behavior, dialogue
  management module)
• Moses Hall (interface, integration,
  implementation of system modules)
• Min Chi (experienced dialogue author)

27
More info software download

• http//andes2.lrdc.pitt.edu/TuTalk
• pjordan_at_pitt.edu
• mosesh_at_pitt.edu

28
Demo snapshots
29
Authoring a dialogue with subdialogues
30
Authoring a subdialogue
31
Extending concept definitions
32
Previewing authored dialogues
33
Testing with dialogue agent server