LTFLL > ALT-C, Nottingham, England, 7-9 September 2010

Alt-C Nottingham, England, 7-9 September 2010

Gaston Burek, Gillian Armitt, Isobel Braidman, Dale Gerdemann, Bernhard Hoisl, Robert Koblischke, Christoph Mauerhofer, Petya Osenova, Kiril Simov, Alisdair Smithies and Fridolin Wild
Download Poster

Two Language Technology-Based Services for Providing Personalised Formative Feedback

Background:
Learners can benefit from personalised formative feedback while learning but this is often not practical owing to tutors' availability and time constraints.  This demonstration provides two examples from the EU-funded LTfLL project of use of language technologies to provide feedback on learner texts.  Both services implement latent semantic analysis (LSA) (Landauer and Dumais 1997), which uses statistical computations to analyse textual relationships and facilitate comparisons between learner and reference texts.  The theoretical framework for both scenarios is Stahl’s model of knowledge building (Stahl 2006), which integrates personal and collaborative “knowing”.
Scenario 1:
WP4.1 is aimed at learners where expected knowledge has boundaries, e.g. in the early stages of a learning journey.  Feedback is situated within a positioning service that poses questions to be answered. Learners may know more than they actually articulate, so feedback helps them revise their texts before submission to tutors for positioning.
Scenario 2:
CONSPECT is aimed at advanced learners with high degrees of autonomy regarding what they learn and when.  Learners can find it difficult to determine the required depth and extent of knowledge.  CONSPECT compares learner texts with those of peers and reference materials to identify gaps in conceptual coverage.


Approach:
Scenario 1:
WP4.1 combines linguistic knowledge and concept coverage. Learner texts are compared with expert texts by means of identifying distinctive phrases and linguistic patterns and implementing LSA (Burek and Gerdemann 2009). Ontology and lexical resources are used to identify concepts covered by learners. A commercial training company is piloting the software for short introductory IT courses.
Scenario 2:
LSA is used to compare learners' texts semantically with reference texts, e.g. from peers, tutors or learning materials.  Force direction, an optimised graph layout algorithm (Fruchterman and Reingold 1991), underpins an interactive two dimensional visualisation of the LSA output (conceptogram).  Users interact with conceptograms to identify future areas of learning. 

Session Activities:
Introductions to language technologies and scenarios.
Hands-on activities:
Scenario 1:
- Learners respond to a question and use feedback from the software to revise their text.
- Tutors see and adjust feedback and provisional grading.
Scenario 2:
Advanced learners:
- input text and explore the resulting conceptogram.
- compare their conceptogram with those of other users to identify areas for further study.
Feedback

Intended outcomes:
• Greater understanding of the potential and limitations of language technologies for providing formative feedback.
• Insight into possible changes in learning approaches.

References
Burek, G.G., and D. Gerdemann. 2009. Maximal Phrases Based Analysis for Prototyping Online Discussion Forums Postings. In Proceedings of the workshop on Adaptation of Language Resources and Technology to New Domains (AdaptLRTtoND),  RANLP Conference, Borovets, Bulgaria, 17 September, 2009.
 
Fruchterman, T., and E. Reingold. 1991. Graph Drawing by Force-directed Placement.  Software - Practice and Experience 21: 1129-1164.

Landauer, T.K., and S. Dumais. 1997. A Solution to Plato's Problem: The latent semantic analysis theory of acquisition, induction and representation of knowledge. Psychological Review 104: 211-240.

Stahl, G. 2006. Group Cognition: Computer Support for Building Collaborative Knowledge. Cambridge: MIT Press.

Indicative timings:
Introduction: 10 minutes
Hands-on activity: 30 minutes
Feedback: 10 minutes