Jiwon Jason Choi 최지원

BIO

Hello World! I'm Jiwon Jason Choi :)
I'm master’s student at Department of Computer Science and Engineering, Sungkyunkwan University. I'm working for SKKU IDCLab since 2021.05 advised by Jaemin Jo.
My research aims to automate data analysis and presentation to improve data accessibility. This includes: 1) Representation learning that reflects the semantics of tabular data; 2) Multimodal generation models for authoring captions and visualizations; 3) Interactive interfaces that connect data visualization with natural language captions.

CONTACT

Phone
(+82) 10-2056-9216
Blog
Email
Github

RESEARCH INTERESTS

Information Visualization
Human-Computer(AI, Data) Interaction
Controllable Natural Language Generation
Multi-Modal Representation Learning

EXPERIENCES

Researcher
2021.05 ~ Current
Research Project
Automated data analysis & presentation using neural network (visualization, natural language)
Related Paper: Intentable [P1]
Human-in-the-loop AutoML
Related Paper: VANAS [P2]
Teaching Assistant
2022.09 ~ Current
Human-Computer Interaction
2022 Fall
Open Source Software Practice
2022 Fall

EDUCATION

Sungkyunkwan University
2022.09 ~ Current
Master’s Student in Computer Science and Engineering
Sungkyunkwan University
2019.03 ~ 2022.08
B.Sc. in Computer Science Education
Kyunggi Highschool
2015.03 ~ 2018.02
Peer Lecturer in School Computer Science Club

PUBLICATION

[P1] Intentable: A Mixed-Initiative System for Intent-Based Chart Captioning

Jiwon Choi and Jaemin Jo
Proceedings of Conference on 2022 IEEE Visualization & Visual Analytics (IEEE VIS)
We present Intentable, a mixed-initiative caption authoring system that allows the author to steer an automatic caption generation process to reflect their intents on captions. We first derive grammar for specifying the intents, i.e., a caption recipe, and build a neural network that generates caption sentences given a recipe. Our quantitative evaluation revealed that our intent-based generation system not only allows the author to engage in the generation process but also produces more fluent captions than the previous end-to-end approaches without user intervention. Finally, we demonstrate the versatility of our system such as context adaptation, unit conversion, and sentence reordering.

[P2] VANAS: A Visual Analytics System for Neural Architecture Search

Jiwon Choi, Gwon Hong, and Jaemin Jo
Proceedings of HCI KOREA 2022
In this paper, we present VANAS, a system that analyzes and visualizes the results of the Neural Architecture Search (NAS) algorithm. First, We devised several algorithms to quantify the significance and contribution of edges, and to recommend neural architecture to users, in order to efficiently convey the vast search results. We created an overview panel to check the overall characteristics of the search space, as well as a user interface to design and analyze the neural network, based on our algorithms. We demonstrated that by analyzing NAS-Bench-101 using VANAS, users can check the importance and significance of nodes and edges that compose neural networks, and that the accuracy can be increased by adding edges recommended by VANAS.

SKILL

Programming Language
Python, JavaScript / TypeScript, C, Java
Database
MySQL(MariaDB), MongoDB
Framework and Library
React, Next.js, FastAPI, PyTorch, HuggingFace
Visualization
d3.js, visx, Vega-Lite
ETC
AWS, Docker, NGiNX, Framer, Figma, Illustrator