WikiTableEdit: A Benchmark for Table Editing by Natural Language Instruction
This work addresses a domain-specific problem for researchers and practitioners dealing with complex tabular data, but it is incremental as it builds on existing datasets and focuses on a niche aspect of table editing.
The paper tackles the challenge of editing irregular tables with merged cells using natural language instructions by introducing the WikiTableEdit dataset, which includes over 200,000 instances generated from 26,531 tables, and evaluates several large language models to show the difficulty of this task.
Tabular data, as a crucial form of data representation, exists in diverse formats on the Web. When confronted with complex and irregular tables, manual modification becomes a laborious task. This paper investigates the performance of Large Language Models (LLMs) in the context of table editing tasks. Existing research mainly focuses on regular-shaped tables, wherein instructions are used to generate code in SQL, Python, or Excel Office-script for manipulating the tables. Nevertheless, editing tables with irregular structures, particularly those containing merged cells spanning multiple rows, poses a challenge when using code. To address this, we introduce the WikiTableEdit dataset. Leveraging 26,531 tables from the WikiSQL dataset, we automatically generate natural language instructions for six distinct basic operations and the corresponding outcomes, resulting in over 200,000 instances. Subsequently, we evaluate several representative large language models on the WikiTableEdit dataset to demonstrate the challenge of this task. The dataset will be released to the community to promote related researches.