A5021946392- Software Engineering Research
- Topic Modeling
- Design Education and Practice
- Personal Information Management and User Behavior
- Speech and dialogue systems
- Data Visualization and Analytics
- AI in cancer detection
- Image Retrieval and Classification Techniques
- Biomedical Text Mining and Ontologies
- Brain Tumor Detection and Classification
- Multimedia Communication and Technology
- Innovative Human-Technology Interaction
- Creativity in Education and Neuroscience
- Software Engineering Techniques and Practices
- Scientific Computing and Data Management
- Usability and User Interface Design
- Digital Games and Media
- AI in Service Interactions
- Explainable Artificial Intelligence (XAI)
- Video Analysis and Summarization
- Persona Design and Applications
Korea Advanced Institute of Science and Technology
2023-2024
By simply composing prompts, developers can prototype novel generative applications with Large Language Models (LLMs). To refine prototypes into products, however, must iteratively revise prompts by evaluating outputs to diagnose weaknesses. Formative interviews (N=8) revealed that invest significant effort in manually as they assess context-specific and subjective criteria. We present EvalLM, an interactive system for refining multiple on user-defined describing criteria natural language,...
Large Language Models (LLMs) have become the backbone of numerous writing interfaces with goal supporting end-users across diverse tasks. While LLMs reduce effort manual writing, may need to experiment and iterate various generation configurations (e.g., inputs model parameters) until results meet their goals. However, these are not designed for experimentation iteration, can restrict how track, compare, combine configurations. In this work, we present "cells, generators, lenses", a...
Children acquire an understanding of the world by asking "why" and "how" questions. Conversational agents (CAs) like smart speakers or voice assistants can be promising respondents to children's questions as they are more readily available than parents teachers. However, CAs' answers not designed for children, difficult understand provide little interactivity engage child. In this work, we propose design guidelines creating interactive dialogues that promote engagement help them...
Designers rely on visual search to explore and develop ideas in early design stages. However, designers can struggle identify suitable text queries initiate a or discover images for similarity-based that adequately express their intent. We propose GenQuery, novel system integrates generative models into the process. GenQuery automatically elaborate users' surface concrete directions when users only have abstract ideas. To support precise expression of intents, enables generatively modify use...
Research consumption has been traditionally limited to the reading of academic papers-a static, dense, and formally written format. Alternatively, pre-recorded conference presentation videos, which are more dynamic, concise, colloquial, have recently become widely available but potentially under-utilized. In this work, we explore design space benefits for combining papers talk videos leverage their complementary nature provide a rich fluid research experience. Based on formative co-design...
Despite the growing demand for professional graphic design knowledge, tacit nature of inhibits knowledge sharing. However, there is a limited understanding on characteristics and instances in design. In this work, we build comprehensive set through literature review. Through interviews with 10 designers, collected 123 labeled their characteristics. By qualitatively coding instances, identified prominent elements, actions, purposes knowledge. To identify which have been addressed least,...
As Large Language Models (LLMs) are nondeterministic, the same input can generate different outputs, some of which may be incorrect or hallucinated. If run again, LLM correct itself and produce answer. Unfortunately, most LLM-powered systems resort to single results which, not, users accept. Having multiple outputs help identify disagreements alternatives. However, it is not obvious how user will interpret conflicts inconsistencies. To this end, we investigate perceive AI model comprehend...
When developing Computer Aided Detection (CAD) systems for Digital Breast Tomosynthesis (DBT), the complexity arising from volumetric nature of modality poses significant technical challenges obtaining large-scale accurate annotations. Without access to annotations, resulting model may not generalize different domains. Given costly DBT how effectively increase amount data used training CAD remains an open challenge. In this paper, we present SelectiveKD, a semi-supervised learning framework...