metadata management

This study addresses the low accuracy of large language models (LLMs) in FAIR-compliance validation of biosample metadata. We propose a structured-knowledge-guided prompting method, integrating the CEDAR template repository, domain-specific data dictionaries, and GPT-4 to construct a metadata standards-conformance verification framework—demonstrated on human lung cancer biosamples. Experimental results show that incorporating structured knowledge significantly improves field-level standards compliance from 79% to 97% (p < 0.01), providing the first empirical evidence that structured knowledge bases can overcome performance bottlenecks inherent to purely text-based LLM prompting in metadata governance. Our approach establishes a novel paradigm for automated, high-accuracy, and interpretable FAIR metadata quality control, enabling scalable, standards-aware curation of biomedical metadata.

To address insufficient standardization, poor domain adaptability, and practical challenges in implementing FAIR principles for research data metadata in open science, this paper proposes a knowledge engineering–based metadata templating approach. It formalizes domain-specific metadata standards as reusable, logically inferable knowledge templates; designs lightweight, dual-mode (Web form and spreadsheet) acquisition interfaces with real-time validation; and enables cross-platform intelligent integration via declarative knowledge representation. The method has been adopted by multiple international scientific consortia to establish standardized metadata frameworks and has driven the development of over ten data annotation systems. Empirical outcomes demonstrate significant improvements in metadata syntactic and semantic consistency, domain alignment, and cross-platform interoperability. By embedding domain knowledge into machine-processable templates, the approach delivers a scalable, maintainable, knowledge-driven paradigm for FAIR data infrastructure.

Scientific data often suffers from poor discoverability, limited sharing, inefficient reuse, and high curation costs due to inadequate standardization of fields and terminology. To address these challenges, this paper proposes a FAIR-aligned standardization framework. Methodologically, it integrates structured vocabulary design, context-aware metadata modeling, and data homogenization strategies to systematically mitigate semantic noise and concept explosion. Crucially, it embeds ten actionable, principle-based rules into a dynamic, evolving data governance process. Empirical evaluation demonstrates that the framework significantly improves metadata quality and semantic consistency, reduces data management overhead, and enhances data findability, interoperability, and long-term reusability—thereby enabling robust, real-world implementation of the FAIR principles in scientific research settings.

To address insufficient semantic description of multidimensional aggregate/summary data, poor adaptability of metadata standards, and cross-source interoperability challenges in big data environments, this paper proposes a multidimensional data source profiling metadata model tailored for data ecosystems. Built upon RDF, the model is the first to support extensible semantic modeling of both aggregate and summary multidimensional data, enabling semantic alignment of dimensions and measures with reference knowledge graphs. It integrates multi-granularity metadata profiles—spanning source-level, attribute-level, and value-distribution characteristics. The model ensures flexible extensibility and cross-source interoperability. Experimental results demonstrate that profile generation time scales linearly with data cardinality, confirming its engineering practicality and predictable performance.

Although scientific data increasingly adhere to the FAIR principles and employ standardized identifiers, practical interoperability remains hindered by heterogeneity in identifier systems and data models. This work proposes and implements two synergistic tools—Babel and ORION—to bridge this gap. Babel constructs clusters of equivalent identifiers through mapping-based clustering and exposes them via a high-performance quantitative API, while ORION standardizes heterogeneous knowledge bases by aligning them to a community-governed common data model. Together, they systematically address the longstanding disconnect between the FAIR “Interoperable” principle and its real-world implementation. The integration of these tools has enabled the construction of a fully interoperable knowledge base, substantially enhancing cross-resource data integration and query capabilities. The resulting framework is publicly available.

This study addresses the challenge in attributed graph schema design of whether repeatedly occurring descriptive attributes should be embedded within nodes or externalized as reusable metadata. Building upon Fifth Normal Form (5NF), the authors propose a principled decision framework that systematically identifies metadata candidates based on semantic criteria rather than mere repetition frequency. The approach classifies attributes into characteristic nodes, embedded properties, or borderline cases using five key principles: cross-element occurrence frequency, conceptual independence, lossless externalizability, reuse potential, and governance relevance. Empirical validation through a library domain case study and an entity classification task demonstrates that repetition alone is insufficient for externalization decisions—semantic judgment is essential. The proposed method significantly enhances the accuracy, consistency, and reusability of metadata modeling in graph-based systems.

This study addresses the lack of systematic guidance on contextualization strategies for large language model (LLM) agents operating in structured data environments, particularly concerning effectiveness and efficiency across multi-file, large-scale schemas. Using SQL generation as a proxy task, the work presents the first systematic evaluation of eleven models across four context formats—YAML, Markdown, JSON, and TOON—at schema scales ranging from 10 to 10,000 tables. The findings reveal that model capability tiers critically determine optimal context architecture: tailored strategies significantly improve performance, with state-of-the-art models gaining 2.7% accuracy under native file-based contexts, while open-source models average a 7.7% decline. Moreover, native file-based agents scale efficiently to ten-thousand-table schemas while maintaining high navigation accuracy.

This study investigates the potential of non-URL metadata to accelerate large language model (LLM) pretraining and explores effective mechanisms for its integration. Method: We propose a novel metadata-postposition paradigm and a learnable metadata tokenization approach, jointly optimizing with auxiliary prediction tasks and masked loss. Metadata types, injection positions, and modeling strategies are systematically examined. Contribution/Results: We present the first systematic empirical validation of training acceleration using fine-grained document quality signals and other metadata categories. Representation probing reveals that metadata significantly reshapes latent representation structures, with fine-grained information encoding playing a critical role. Experiments demonstrate substantial improvements in pretraining efficiency—up to 1.8× faster convergence on downstream benchmarks—while maintaining or improving model quality. Our framework provides a reusable, quality-aware, structured modeling architecture for efficient LLM training, generalizable across diverse metadata modalities and model scales.

This study addresses the limitations of semantic similarity–based retrieval in structured, highly repetitive regulatory texts, where linguistic overlap often obscures meaningful content distinctions and undermines the effectiveness of retrieval-augmented generation (RAG). To mitigate this issue, the work systematically investigates metadata-aware retrieval strategies, proposing and evaluating fusion approaches such as unified embedding and prefix concatenation. The findings demonstrate that incorporating metadata enhances intra-document cohesion and reduces inter-document ambiguity, thereby improving retrieval performance. Evaluated on a newly curated benchmark dataset, RAGMATE-10K, both the unified embedding and prefix-based methods significantly outperform pure text baselines across multiple question types and evaluation metrics. Notably, the unified embedding approach achieves superior performance while maintaining greater maintainability.

Keyword search in open data portals suffers from incomplete and inconsistent metadata, as well as a semantic gap between user queries and domain-specific terminology, leading to poor natural language query performance. This work focuses on the critical role of metadata fields—particularly descriptive text—in enabling semantic alignment and proposes a large language model (LLM)-based method to automatically generate high-quality dataset descriptions that bridge the gap between user intent and structured metadata. Through controlled ablation studies and simulated natural language queries, we systematically evaluate retrieval performance across varying metadata configurations and prompting strategies. Results demonstrate that LLM-generated descriptions significantly improve both precision and recall, validating the effectiveness and practicality of generative approaches for enhancing open data discoverability. Our approach establishes a reproducible, scalable paradigm for metadata enrichment in open data ecosystems.