Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLMClean: Context-Aware Tabular Data Cleaning via LLM-Generated OFDs

Fabian Biester, Mohamed Abdelaal, Daniel Del Gaudio

TL;DR

LLMClean automates the generation of context models for OFD-based data cleaning by leveraging LLMs to analyze real-world datasets, reducing the manual effort required from domain experts. It distinguishes IoT and non-IoT data, builds a meta-context model via a three-stage pipeline, and outputs OFDs that guide error detection and repair. Through prompt ensembling and extensive experiments on IoT, Hospital, and CONTEXT datasets, LLMClean demonstrates competitive cleaning performance against baseline tools and shows robust handling of diverse error patterns. The work advances data quality for AI pipelines by providing scalable, automated context modeling and integrating with RDF-based semantic representations, with practical implications for Industry 4.0 and IoT deployments.

Abstract

Machine learning's influence is expanding rapidly, now integral to decision-making processes from corporate strategy to the advancements in Industry 4.0. The efficacy of Artificial Intelligence broadly hinges on the caliber of data used during its training phase; optimal performance is tied to exceptional data quality. Data cleaning tools, particularly those that exploit functional dependencies within ontological frameworks or context models, are instrumental in augmenting data quality. Nevertheless, crafting these context models is a demanding task, both in terms of resources and expertise, often necessitating specialized knowledge from domain experts. In light of these challenges, this paper introduces an innovative approach, called LLMClean, for the automated generation of context models, utilizing Large Language Models to analyze and understand various datasets. LLMClean encompasses a sequence of actions, starting with categorizing the dataset, extracting or mapping relevant models, and ultimately synthesizing the context model. To demonstrate its potential, we have developed and tested a prototype that applies our approach to three distinct datasets from the Internet of Things, healthcare, and Industry 4.0 sectors. The results of our evaluation indicate that our automated approach can achieve data cleaning efficacy comparable with that of context models crafted by human experts.

LLMClean: Context-Aware Tabular Data Cleaning via LLM-Generated OFDs

TL;DR

LLMClean automates the generation of context models for OFD-based data cleaning by leveraging LLMs to analyze real-world datasets, reducing the manual effort required from domain experts. It distinguishes IoT and non-IoT data, builds a meta-context model via a three-stage pipeline, and outputs OFDs that guide error detection and repair. Through prompt ensembling and extensive experiments on IoT, Hospital, and CONTEXT datasets, LLMClean demonstrates competitive cleaning performance against baseline tools and shows robust handling of diverse error patterns. The work advances data quality for AI pipelines by providing scalable, automated context modeling and integrating with RDF-based semantic representations, with practical implications for Industry 4.0 and IoT deployments.

Abstract

Machine learning's influence is expanding rapidly, now integral to decision-making processes from corporate strategy to the advancements in Industry 4.0. The efficacy of Artificial Intelligence broadly hinges on the caliber of data used during its training phase; optimal performance is tied to exceptional data quality. Data cleaning tools, particularly those that exploit functional dependencies within ontological frameworks or context models, are instrumental in augmenting data quality. Nevertheless, crafting these context models is a demanding task, both in terms of resources and expertise, often necessitating specialized knowledge from domain experts. In light of these challenges, this paper introduces an innovative approach, called LLMClean, for the automated generation of context models, utilizing Large Language Models to analyze and understand various datasets. LLMClean encompasses a sequence of actions, starting with categorizing the dataset, extracting or mapping relevant models, and ultimately synthesizing the context model. To demonstrate its potential, we have developed and tested a prototype that applies our approach to three distinct datasets from the Internet of Things, healthcare, and Industry 4.0 sectors. The results of our evaluation indicate that our automated approach can achieve data cleaning efficacy comparable with that of context models crafted by human experts.
Paper Structure (28 sections, 10 figures, 2 tables, 2 algorithms)

This paper contains 28 sections, 10 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Overview
  3. OFD Dependencies
  4. Datasets Categorization
  5. Automated Context Modeling
  6. Handling IoT Datasets
  7. Sensor Splitting
  8. Column Generation
  9. Column Renaming
  10. Handling Non-IoT Datasets
  11. Prompt Ensembling
  12. Data Cleaning with LLMClean
  13. Performance Evaluation
  14. Experimental Setup
  15. LLM Models
  16. ...and 13 more sections

Figures (10)

  • Figure 1: Architecture of LLMClean
  • Figure 2: Metamodel of the context model for all IoT datasets
  • Figure 3: Automated generation of context model
  • Figure 4: Prompt ensembling method
  • Figure 5: Impact of few-shot learning, where BE denotes the best ensemble and BP denotes the best prompt
  • ...and 5 more figures