dataset preprocessing

In the context of open science, there is an urgent need to implement the FAIR principles (Findable, Accessible, Interoperable, Reusable) in scientific data management, yet systematic guidance on achieving FAIR compliance through big data software reference architectures (SRAs) remains lacking. Method: We conducted a rigorous systematic literature review, screening 323 publications—including those from authoritative databases and expert recommendations—and performed structured data extraction and evaluation aligned with predefined research questions. Contribution/Results: The study identifies seven generic FAIR-compliant SRAs, thirteen scenario-specific FAIR pipelines, and three fully FAIR-compatible SRAs. It uncovers critical bottlenecks in metadata standardization, cross-platform interoperability, and long-term reusability. Furthermore, it establishes the first classification framework and empirical evaluation system for FAIR-oriented big data SRAs, thereby filling a significant research gap and providing a methodological foundation and strategic direction for future SRA design, policy formulation, and tool development.

A significant gap exists between theoretical definitions of data quality dimensions—such as accuracy, completeness, consistency, and timeliness—as stipulated in standards (e.g., ISO/IEC 25012) and the actual functionalities implemented in widely used data quality tools, with no systematic mapping analysis to date. Method: We conducted a systematic literature review and performed functional reverse engineering on seven mainstream open-source data quality tools. Contribution/Results: We present the first many-to-many mapping framework linking data quality dimensions to concrete tool capabilities, introducing a cross-dimensional, fine-grained functional categorization schema and generating a structured correspondence matrix covering all core dimensions. This work bridges the theory–practice divide, delivers an actionable guide for data quality assessment, and substantially enhances the scientific rigor and reusability of tool selection, functionality design, and standard implementation.

This study addresses a critical gap in data cleaning research—the lack of large-scale, real-world dirty datasets that hinder the effective evaluation of methods in practical settings. To bridge this gap, the authors construct the first large-scale dirty dataset comprising real postal addresses paired with their ground-truth counterparts. Leveraging this dataset, they conduct a systematic benchmarking study of state-of-the-art data cleaning approaches. Their experiments reveal substantial limitations of current methods when applied to real-world data, underscoring the need for more robust and context-aware techniques. The dataset and empirical findings not only establish a reliable benchmark for future research but also provide actionable insights to guide the development of data cleaning solutions tailored to real-world applications.

Existing data management tools suffer from limited automation, poor interactivity, and insufficient integration with ML workflows, compromising data quality and hindering analytical and modeling performance. To address this, we propose an interactive, ML-oriented tabular data quality dashboard that establishes an adaptive, human-in-the-loop + ML-driven data cleaning闭环. Our approach integrates data profiling, multi-strategy error detection and repair—including statistical analysis, rule-based engines, and supervised/semi-supervised models—while supporting expert rule validation and labeling. Cleaning strategies are iteratively refined using downstream model performance feedback. Furthermore, we unify DataSheets, MLflow, and Delta Lake to ensure reproducibility, traceability, and versioning of the cleaning pipeline. Experiments across multiple benchmark datasets demonstrate significant improvements: error identification rate and repair accuracy increase notably, downstream ML models achieve an average 7.2% accuracy gain, and cleaning time decreases by 40%.

Existing fairness-aware machine learning research suffers from limited dataset availability, ad hoc dataset selection, inconsistent preprocessing, and insufficient metadata—resulting in poor generalizability and low reproducibility. To address these challenges, we introduce FairGround: the first unified data framework specifically designed for algorithmic fairness research. FairGround comprises 44 diverse, tabular datasets spanning multiple domains, each richly annotated with fine-grained fairness-related metadata—including formal definitions of sensitive attributes, bias types, and socio-contextual information. Complementing the framework is an open-source Python toolkit that enables standardized data loading, configurable preprocessing, principled train/validation/test splits, and end-to-end reproducible experimentation. By providing consistent, transparent, and well-documented resources, FairGround significantly enhances the consistency, comparability, and reproducibility of fairness evaluations. It thereby advances fair machine learning toward greater methodological rigor, transparency, and standardization.

Data standardization is critical in the data science lifecycle, yet existing tools (e.g., Pandas) require manual, error-prone coding, while LLM-based automation still demands expert prompt engineering and iterative interaction. To address this, we propose a declarative API-driven LLM-Agent framework that introduces *Dataprep.Clean*—a novel, column-type-aware standardization component enabling end-to-end automation via a single-line operation and one-shot natural language input. Our method integrates domain knowledge modeling with lightweight agent orchestration, eliminating programming prerequisites and enabling semantic cleaning of heterogeneous columns. Evaluated on real-world datasets, the approach achieves high accuracy and robustness across diverse standardization tasks. Deployed as an interactive web tool, it substantially lowers the barrier to entry for data practitioners. This work advances data preprocessing toward declarative, intelligent automation—bridging the gap between domain expertise and scalable, user-friendly tooling.

Data cleaning remains highly manual, inefficient, and error-prone. This paper proposes the first goal-driven LLM-based framework for automatic workflow generation: given a dirty table and a target query, it end-to-end generates a minimal viable clean table along with executable cleaning steps—including deduplication, missing-value imputation, and format standardization. Our contributions are threefold: (1) We introduce the first benchmark dataset comprising annotated quadruples of (goal, dirty table, cleaning workflow, cleaned answer); (2) We design a zero-shot, multi-stage prompting framework—requiring no fine-tuning—that decomposes the task into goal column identification, data quality diagnosis, and operation-parameter generation; (3) We empirically validate that off-the-shelf LLMs possess inherent reasoning capabilities sufficient to generate high-quality, executable cleaning workflows across three major LLM families, significantly reducing human intervention.

This work addresses common data quality issues in tabular data—such as missing values, inconsistent formatting, dependency violations, unit errors, and categorical ambiguities—by proposing a training-free cleaning approach. The method leverages few-shot annotations to guide large language models (LLMs) in reasoning about data flaws and automatically synthesizing reusable Python cleaning scripts equipped with guard conditions. Central to this approach is an evidence-based guarded repair mechanism that applies deterministic transformations only when specific dirty data patterns are detected and sufficiently supported by contextual evidence. By integrating data profiling, program synthesis, and cell-level validation, the framework ensures high logical precision. Evaluated on six benchmark datasets, the method achieves higher F1 scores than state-of-the-art rule-based, learning-based, and LLM-driven baselines on five datasets, while substantially reducing computational overhead and LLM invocation costs during repeated cleaning runs.

Traditional static data cleaning methods exhibit inconsistency and poor adaptability when applied to dynamic data streams—such as continuous temporal sensor measurements (e.g., temperature, illumination)—posing significant challenges for real-time error detection and correction. Method: This work identifies the inherent non-stationarity of stream cleaning and establishes the first theoretical framework specifically designed for streaming environments, formally characterizing its unique challenges: timeliness constraints, state evolution, and incremental validation. Based on this framework, we design a prototype system and conduct empirical evaluations across diverse streaming scenarios. Contribution/Results: Our analysis reveals substantial performance volatility in existing approaches, pinpointing three critical factors: windowing strategy, error propagation patterns, and data arrival rate. The framework provides a scalable theoretical foundation and a rigorous evaluation methodology for stream data cleaning, enabling principled design and comparative assessment of future techniques.

This work addresses the gap between theory and practice in error detection and cleaning for tabular data by proposing and implementing an interactive web-based demonstration system that integrates error modeling, injection, and intelligent cleaning. For the first time, the system unifies machine learning–based data cleaning methods with dependency-aware error generation models within a visual platform, enabling users to upload tables, inject realistic errors, and interactively experience the cleaning process alongside interpretable explanations of the underlying mechanisms. By synergizing machine learning, database technologies, and web development, the project enhances intuitive understanding of error repair principles and provides a publicly accessible demo platform (https://cured.demo.calgo-lab.de/) that serves as an effective validation tool for both research and practical applications in data cleaning.

Data preparation for AI applications in materials discovery, molecular modeling, and climate science remains labor-intensive, expert-dependent, and poorly scalable. Method: This paper introduces the first end-to-end autonomous data agent system, featuring dual feedback loops that integrate large language models’ (LLMs) semantic understanding with domain-knowledge-guided grounded validation. It employs hierarchical routing, automated cleaning and normalization, and feature-level optimization algorithms to transform raw tabular data into AI-ready datasets without human intervention. Contribution/Results: Evaluated on real-world datasets across multiple scientific domains, the system significantly improves data quality and downstream model performance. It establishes the first safe, reliable, and non-expert-friendly paradigm for autonomous data engineering—enabling fully automated, robust, and interpretable data curation tailored to domain-specific requirements.

Existing synthetic data generation tools often suffer from complex workflows, inconsistent standards, and limited cross-modal extensibility, hindering their ability to meet the data demands of large language models in specialized domains and low-resource languages. This work proposes a configuration-driven, end-to-end open-source framework that standardizes multi-source data synthesis through a unified and controllable paradigm. Featuring a highly modular architecture, the framework flexibly adapts to diverse tasks and supports high-quality data generation across multiple pathways, modalities, and languages. Integrated with both a graphical user interface and command-line utilities, it significantly lowers the barrier to entry for users. Empirical evaluations demonstrate that the framework effectively balances generation efficiency and data quality across various scenarios, thereby accelerating the practical deployment of synthetic data in model training pipelines.