Improving Table Understanding with LLMs and Entity-Oriented Search

TL;DR

This paper introduces an entity-oriented search method combined with a graph query language to enhance large language models' understanding of tables, achieving state-of-the-art results on key benchmarks.

Contribution

It presents a novel entity-oriented search approach and a graph query language for improved table understanding with LLMs, reducing preprocessing needs and enhancing contextual reasoning.

Findings

Achieved state-of-the-art performance on WikiTableQuestions

Outperformed existing methods on TabFact benchmark

Reduced reliance on data preprocessing and keyword matching

Abstract

Our work addresses the challenges of understanding tables. Existing methods often struggle with the unpredictable nature of table content, leading to a reliance on preprocessing and keyword matching. They also face limitations due to the lack of contextual information, which complicates the reasoning processes of large language models (LLMs). To overcome these challenges, we introduce an entity-oriented search method to improve table understanding with LLMs. This approach effectively leverages the semantic similarities between questions and table data, as well as the implicit relationships between table cells, minimizing the need for data preprocessing and keyword matching. Additionally, it focuses on table entities, ensuring that table cells are semantically tightly bound, thereby enhancing contextual clarity. Furthermore, we pioneer the use of a graph query language for table understanding, establishing a new research direction. Experiments show that our approach achieves new state-of-the-art performances on standard benchmarks WikiTableQuestions and TabFact.