Towards an AI-Driven and Intuitive Chat Interface for Generating Visualizations

As data analytics and visualization become indispensable across diverse fields, the complexity of datasets often hinders non-experts from effectively exploring and interpreting crucial insights. Yet, advances in artificial intelligence (AI) techniques, especially natural language processing (NLP), and visual analytics present an opportunity to close this gap by providing conversational approaches to data exploration. The goal of this research is to design, develop, and evaluate an AI-driven and intuitive chat interface for generating visualizations within the Engage Nova Scotia’s Quality of Life Analytics System. We lay the groundwork for achieving this goal by investigating how visual context and scaffolding can be integrated with a natural language interface such that individuals without detailed knowledge of the underlying dataset can generate useful visualizations and correctly interpret them. Specifically, we will design functional prototypes of chat interfaces showcasing different visual context and scaffolding strategies (e.g., interactive overviews, dynamic visual cues, progressive disclosure, etc.), conduct a comparative user study to assess the interfaces’ effectiveness, usability, and user experience, and offer practical recommendations based on evaluation findings. This work contributes to the democratization of data exploration by reducing technical and knowledge barriers for interactively generating insightful visualizations.

Faculty Supervisor:

Oladapo Oyebode;Derek Reilly

Student:

Partner:

Engage Nova Scotia

Discipline:

Computer science

Sector:

Health and Related Sciences & Technology; Professional, scientific and technical services

University:

Dalhousie University

Program: