Touching other people's work

1. User modeling
2. L2 Assessment
3. Recommender Systems
4. Crowdsourcing

User modeling

UM is ubiquitous

Learner modeling as subset of user modeling

Most learner modeling in programming

What about language learners?

L2 Assessment

Static vs. dynamic

In language learning, assessment as knowledge at point in time

What about continuous assessment?

Recommender Systems

What to do next?

What to do next?

Based on user model

What to do next?

Based on user model

or crowd model

What to do next?

Based on user model

or crowd model

"People who liked this product also liked this product"

Applicability to language learning?

Crowdsourcing

Collection of data through the masses

What crowd?

Why

Language learning individual experience

Learner differences

• Prior knowledge
• Learning speed
• Learning style
• Language background

Online learning platforms rarely personalized

Fixed progression path

Personalized learning environments

Collection of available tools

How

Collect data from learners using the platform

Profile: static

Learner model: dynamic

• Track learner activity on platform
• Continuous assessment
• Stealth assessment

Roadmap

• 2016: Requirements for exercises
• 2017:
  • Lärka exercises set up
  • Diagnostic test
• 2017 1/2: Data collection and evaluation
• 2018: Data description/evaluation #1
• Mittseminarium: Theoretical user model
• 2019: User modeling in place
• 2019-2020: Study #2
• 2020-2021:
  • Wrap up
  • Writing

What has been done

• Ildikó Pilán, David Alfter, Elena Volodina. Coursebook texts as a helping hand for classifying linguistic complexity in language learners' writings. Proceedings of the workshop on Computational Linguistics for Linguistic Complexity (CL4LC), COLING 2016, Osaka, Japan.
• David Alfter, Elena Volodina. Modeling Individual Learner Knowledge in a Computer Assisted Language Learning System. Proceedings of SLTC 2016, Umeå, Sweden.
• David Alfter, Yuri Bizzoni, Anders Agebjörn, Elena Volodina, Ildikó Pilán. From Distributions to Labels: A Lexical Proficiency Analysis using Learner Corpora. Proceedings of the joint workshop on NLP for Computer Assisted Language Learning and NLP for Language Acquisition at SLTC 2016. Linköping Electronic Conference Proceedings 130: 1–7. Umeå, Sweden.
• Elena Volodina, Ildikó Pilán, David Alfter. Classification of Swedish learner essays by CEFR levels. Proceedings of EuroCALL 2016. Limassol, Cyprus.
• David Alfter, Elena Volodina. Learning the Learner: User Modeling in Intelligent Computer Assisted Language Learning Systems. CEUR Workshop proceedings. UMAP 2016, Halifax, Canada.

• David Alfter, Elena Volodina. Adaptive diagnostic test.
• David Alfter, Sandra Morales. Teachers' voices: A reflective approach to explore the use of an online platform for language teaching
• David Alfter. Evaluating sentence rearrangment and sentence composition tasks using Partial Tree Kernels

2017 - upcoming

What is being done

Diagnostic test

Implementation of

• exercises
• login
• logging

What profile data?

In collaboration with SweLL project

Profile

Education background

Language background

Learner model

Data collected during interaction with learning platform

General data

Not learner-specific

• Current time
• Frequency of use
• Time since last visit
• Screen size of device
• IP address
• Geolocation

Multiple choice vocabulary exercise

• Time taken
• History
• Final answer

Variables to collect

Spelling exercise

Spelling exercise

• Frequent (L1 specific) misspellings
• Difficult sounds
• Diagnostic and prognostic

Hypothesis testing

Evaluation of automatically graded vocabulary list

Grading of new vocabulary items

What will be done

• Data collection
• Data analysis
• Learner modeling

What to model?

Phonology

RecSys?

Vocabulary

Difficult

Vocabulary

Focus on...

Grammar

Needs in-depth error analysis

To infinity and beyond

• Morphology
• Grammar

Error analysis

Error groups

Closed tasks

More specific, less rich

Open tasks

Less specific, richer

Feedback?