Title: RealESL ESL Tutoring Using Authentic Materials
1RealESL ESL Tutoring Using Authentic Materials
- Key Ideas
- Authentic materials are different from artificial
materials - E.g., the most common use of shake is shake
hands (6th dictionary entry) - Does practice with artificial materials transfer
to authentic materials? - Authentic materials from a large, open, and
dynamic corpus - Materials that can be used/tailored for a wide
range of very specific instructional needs - Minimize human effort in creating and updating
instructional materials - Automatic or semi-automatic selection, analysis,
annotation, and addition of materials - Using state-of-the-art human language
technologies - Support a variety of student interactions
- E.g., vocabulary-based, grammar-based
- Support a range of language-based assessment
- E.g., multiple choice, cloze, short-answer,
question and short answer
Who We Are Jamie Callan, CMU Robert DeKeyser,
Pitt Maxine Eskenazi, CMU Lori Levin, CMU Alan
Juffs, Pitt Teruko Mitamura, CMU Carolyn Penstein
Rose, CMU Chuck Perfetti, Pitt
LearnLabs Henderson Media Lab (English Language
Institute) Modern Languages
- Significance
- Focus on authentic materials, as opposed to
artificial materials - More opportunity to coordinate tutoring with
other instructional objectives - Avoids the problem of stale instructional
materials - Instructional materials developed on the fly
from an open, dynamic corpus - Makes it easier to test new learning and
instructional hypotheses as they arise - Modify the way materials are selected, analyzed,
annotated or used - Avoid having to modify materials manually
- Extends the lifespan and utility of tutoring
systems as research platforms - Less bound to the initial research agenda
- Preliminary, proof-of-concept effort started in
Fall, 2003 - Reader-specific Lexical Practice for Reading
Comprehension - Funded by U.S. Department of Education
- CMU and the Univ of Pittsburgh (Callan, Eskenazi,
Perfetti)
- Why an SLC?
- Creating a science of language tutoring requires
a diverse group of researchers - Educational psychologists
- Learning scientists
- Language technologists
- and sustained effort
The Dynamic, Open-Corpus Approach to Language
Tutors
Vocabulary stretch 5? 10? reader-specific? tas
k-specific?
Research Questions
Explicit vs. implicit practice
How is component-level practice integrated?
Fluency Cluster Focus
Language Tutors
English As a Second Language
L1 Reading Comprehension
Cognitive Tutor Authoring Tools
French
Chinese
Frame-Based Question Generation
Construction Grammar Patterns
Grammar Error Detection and Feedback
Model of Student Interests
Tutor Components
Student Model
Curriculum Model
Performance Model
Coherence Metrics
Co-reference Resolution
Text Search
Text Categorization
Statistical Language Modeling
Latent Semantic Analysis
Part-of-Speech Tagging The/AT chief/NN of/PRF
these/DT spies/NN is/VBZ the/AT
celebrated/AJ0 Belle/NP Boyd/NP ./.
Dialog Models
Core Technologies
Named-Entity Tagging The chief of these spies is
the celebrated ltPERSONgtBelle Boydlt/PERSONgt