Table Extraction from Texts: A Literature Review

PhD Qualifying Examination


Title: "Table Extraction from Texts: A Literature Review"

by

Miss Tong LI


Abstract:

Table Extraction from Texts is a task that automatically creates structured
tables from unstructured text. This survey organizes the field by first
defining two main scenarios: Local Table Extraction for creating context-
specific tables from a single text, and Global Table Extraction for filling
a shared schema from many texts. We then review and compare the two main
types of methods. End-to-end generation methods use fine-tuned models to
generate tables directly but are often limited to simple, rectangular
structures. Multi-step extraction methods use large language models to
break the workflow into agentic tasks, such as finding informative
elements, planning the schema, and copying the value, which is more
flexible but less controllable. Finally, we summarize the common datasets
used to test these methods and the metrics for evaluating their results.
This survey provides a structured overview of the task, its methods, and
how they are measured, offering a foundation for understanding current
research and guiding future work.


Date:                   Friday, 28 November 2025

Time:                   12:00noon - 2:00pm

Venue:                  Room 5501
                        Lifts 25/26

Committee Members:      Prof. Bo Li (Supervisor/Chairperson)
                        Prof. Lei Chen (Co-supervisor)
                        Dr. Junxian He
                        Prof. Qiong Luo