Exploring Natural Language Interfaces for Data Visualization

PhD Qualifying Examination

Title: "Exploring Natural Language Interfaces for Data Visualization"

by

Mr. Xubang XIONG


Abstract:

Data visualization (DV) plays a crucial role in real-world applications, by
uncovering hidden patterns, simplifying complex data interpretation, and 
supporting decision-making. However, constructing DVs from scratch presents
significant barriers. Non-expert users often lack proficiency in programming
or declarative visualization languages (DVLs) (e.g., Vega-Lite and VQL),
while GUI-based tools (e.g., Microsoft Excel) offer limited customization due
to pre-defined templates. To address these issues, natural language
interfaces for data visualization have been developed, with the core
Text-to-Visualization (Text-to-Vis) problem defined as generating valid DVLs
from natural language queries and corresponding databases. Existing methods
for these interfaces have evolved through three phases: the traditional
phase, the neural network phase, and the foundation language model phase.
This survey systematically examines the development of natural language
interfaces for data visualization, particularly the Text-to-Vis problem. It
is structured to review the traditional DV pipeline, the existing natural
language interfaces for data visualization, some variants of the
Text-to-Vis problem, and conclude with future directions.


Date:                   Monday, 13 October 2025

Time:                   3:00pm - 5:00pm

Venue:                  Room 5501
                        Lifts 25/26

Committee Members:      Prof. Raymond Wong (Supervisor)
                        Prof. Shing-Chi Cheung (Chairperson)
                        Dr. Junxian He