VACode: A Visual Analytics Systems for Learner-Sourced Code in Large-Scale Computer Science Education

VACode is a visual analytics system that allows instructors to analyze learner-sourced code and to assign groups for code-review-based peer-assisted learning. Learner-sourced programming codes are embedded to high-dimensional vector with codeBERT and visualized with t-SNE. Then, student groups are assigned based on Unshift Clustering Algorithms. Instructors can see source-codes' overview in Plot View, filter with Label View and Group View, and check raw source-code with Code View.

Tech Stack

TypeScript, d3.js, Bootstrap
node.js, express, Huggingface, Scikit-learn

My Work

codeBERT를 이용한 코드 임베딩 및 시각화 파이프라인 디자인
Frontend Engineering
Vanilla Typescript만을 이용한 인터페이스 개발 및 컴포넌트 설계
D3를 이용한 Multidimensional Projection Visualization 및 인터랙션 개발