Schema-Driven Information Extraction from Heterogeneous Tables

TL;DR

This paper investigates the use of large language models for cost-effective, schema-driven extraction of structured data from heterogeneous tables across multiple domains, demonstrating competitive performance without task-specific training.

Contribution

The paper introduces a new schema-driven information extraction task and provides a benchmark for evaluating LLMs on diverse tabular data, highlighting their effectiveness and potential for cost-efficient applications.

Findings

LLMs achieve F1 scores from 74.2 to 96.1 on the benchmark.

Cost-efficient extraction is possible without task-specific pipelines.

Distilling compact models reduces API reliance while maintaining performance.

Abstract

In this paper, we explore the question of whether large language models can support cost-efficient information extraction from tables. We introduce schema-driven information extraction, a new task that transforms tabular data into structured records following a human-authored schema. To assess various LLM's capabilities on this task, we present a benchmark comprised of tables from four diverse domains: machine learning papers, chemistry literature, material science journals, and webpages. We use this collection of annotated tables to evaluate the ability of open-source and API-based language models to extract information from tables covering diverse domains and data formats. Our experiments demonstrate that surprisingly competitive performance can be achieved without requiring task-specific pipelines or labels, achieving F1 scores ranging from 74.2 to 96.1, while maintaining cost efficiency. Moreover, through detailed ablation studies and analyses, we investigate the factors contributing to model success and validate the practicality of distilling compact models to reduce API reliance.