Supervisors
Dr. Ahmed Jawad Qureshi, Professor, University of Alberta
Contact: ajqureshi@ualberta.ca
Collaborators
Junxi Shi, University of Alberta
Project Overview
Bloom's Taxonomy is a widely used framework for classifying cognitive processes in education and research. Its cognitive domain categorizes thinking into six levels - Knowledge, Comprehension, Application, Analysis, Synthesis, and Evaluation - and associates specific action verbs with each level. However, the original verb lists are limited and have not been systematically expanded since the taxonomy was first proposed.
In this project, we developed a multi-stage machine learning pipeline to extend these verb lists. The pipeline starts with candidate verb extraction from large text corpora, followed by WordNet-based semantic filtering to identify plausible cognitive verbs. A one-vs-rest classifier trained on sentence embeddings of the original Bloom verbs then scores each candidate verb's alignment with each cognitive domain. Score calibration and threshold-based acceptance criteria determine which new verbs are added to which domain.
The extended verb lists were then applied to a practical research problem: automated coding of think-aloud verbalizations from a comparative study of AR-CAD and a traditional CAD tool. By classifying the verbs participants used during design tasks into Bloom domains, we were able to compare cognitive engagement patterns across the two tools at different stages of the design process, revealing how each tool shapes the way users think during design.
This work has been published as a preprint on Research Square.
Key Contributions
1. ML Pipeline for Verb Extension: Designed a complete pipeline including candidate extraction, semantic filtering, one-vs-rest classification using sentence embeddings, probability calibration, and threshold-based domain assignment for large-scale labeling of new cognitive verbs.
2. Think-Aloud Analysis Workflow: Built a time-resolved analysis workflow that uses transcription, verb extraction, Bloom-domain classification, and normalized progress binning to compare how participants engage cognitively across different design tools.
3. Integration with AR-CAD Research: Applied the extended taxonomy and analysis workflow to a 20-participant comparative study of AR-CAD and a traditional CAD tool, providing empirical evidence of cognitive engagement differences between immersive and traditional design environments.