Pith. sign in

REVIEW 2 major objections 4 minor 28 references

Making every knowledge-graph node a FAIR Digital Object turns graph RAG into accurate, explainable answers for metadata and ontology questions.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 05:24 UTC pith:2C2HXLWT

load-bearing objection Clean engineering demo that FAIRifying GraphRAG nodes helps, but the headline accuracy jump is mostly by construction of the baseline. the 2 major comments →

arxiv 2607.11464 v1 pith:2C2HXLWT submitted 2026-07-13 cs.IR cs.AIcs.CLcs.DB

FAIR GraphRAG: A Retrieval-Augmented Generation Approach for Semantic Data Analysis

Marlena Fl\"uh , Soo-Yon Kim , Carolin Victoria Schneider , Sandra Geisler This is my paper
classification cs.IR cs.AIcs.CLcs.DB
keywords retrieval-augmented generationFAIR data principlesFAIR Digital Objectknowledge graph constructionlarge language modelGraphRAGbiomedical question answeringontology mapping
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that existing graph-based retrieval-augmented generation systems fall short for scientific domains because their underlying knowledge is not systematically made FAIR. By treating every graph node as a FAIR Digital Object that packages data, metadata, a persistent identifier and ontology links, and by letting large language models help build the schema and extract those elements, the resulting FAIR GraphRAG system can answer domain questions more accurately and with full provenance. On a real RNA-sequencing dataset from gastroenterology the approach raises overall question-answering accuracy from roughly 43 percent to 93 percent and makes every answer fully traceable. A sympathetic reader cares because the same pattern can turn any specialized corpus into a queryable, standards-compliant knowledge base without losing the semantic relationships that graphs already capture.

Core claim

When the nodes of a knowledge graph are themselves FAIR Digital Objects, a graph-RAG pipeline that uses an LLM both to construct the graph and to generate database queries produces substantially higher accuracy, coverage and explainability on complex biomedical questions, especially those that require metadata or ontology links, while simultaneously satisfying the FAIR principles for the underlying knowledge resource.

What carries the argument

The FAIR Digital Object node: each graph node packages a core data artifact, rich metadata, a persistent identifier and semantic ontology links, so that every retrieved unit is already findable, accessible, interoperable and reusable.

Load-bearing premise

That simply omitting the FAIR steps from the same construction pipeline yields a fair baseline, and that results on the first eighty rows of one public RNA-seq series plus forty-two hand-written questions will hold for full clinical corpora.

What would settle it

Re-run the identical pipeline and the same forty-two questions on the complete GEO series (or on an independent clinical RNA-seq cohort) and check whether the accuracy gap between FAIR and non-FAIR versions collapses or remains above thirty percentage points.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Any specialized corpus can be turned into a queryable FAIR knowledge graph whose every node already carries its own PID and ontology links.
  • Complex questions that previously failed because metadata or ontology terms were missing become answerable with full provenance.
  • Physicians or domain experts can inspect the exact Cypher query, the returned FDO and its PID, raising trust in the generated answer.
  • The same LLM-assisted construction pipeline can be reused for education or business data without redesigning the retrieval layer.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the FDO node model scales, regulatory audits of AI answers in medicine become simpler because every claim points to a resolvable, licensed digital object.
  • The performance jump on ontology questions suggests that current RAG systems systematically under-use controlled vocabularies; adding them at construction time may be more effective than post-hoc retrieval tricks.
  • Open-source LLMs lag proprietary ones on Cypher generation, so hybrid deployments that keep sensitive data local while using a stronger model for query synthesis may become standard practice.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proposes FAIR GraphRAG, a GraphRAG framework whose knowledge-graph nodes are FAIR Digital Objects (FDOs) that package core data, rich metadata, persistent identifiers, and ontology links. An LLM-assisted pipeline constructs a property-graph schema, extracts entities, FAIRifies them (PID generation, metadata extraction, BioPortal ontology mapping), and stores the resulting FDO graph in Neo4j; natural-language questions are answered by LLM-generated Cypher queries. On a 80-row subset of GEO RNA-seq series GSE280797 the authors report near-complete FAIR-indicator compliance (Table I) and substantially higher QA accuracy, coverage and explainability than an otherwise identical Non-FAIR baseline that simply omits the FAIRification steps (Tables II–III: gpt-4o-mini overall accuracy 92.86 % vs 42.86 %).

Significance. If the central claim holds, the work supplies a concrete, reusable recipe for making the knowledge resources of GraphRAG systems FAIR at the node level rather than only at the whole-graph level. The open-source construction and RAG interfaces, the formal property-graph model of FDOs, and the clean ablation design are genuine engineering contributions that could be adopted in other specialized domains. The co-design with clinicians and the explicit FAIR-indicator checklist further increase practical relevance for biomedical data stewardship.

major comments (2)
  1. [§IV.B.2, Tables II–III] §IV.B.2 and Tables II–III: the headline claim that FAIRification “significantly improves question answering accuracy au particularly for complex queries involving metadata and ontology links” is only weakly isolated. The Non-FAIR baseline (§IV.A.3) is obtained by skipping exactly the PID/metadata/ontology steps of §III.B.3; consequently the 10 metadata and 10 ontology questions receive 0 % accuracy by construction. The residual gain on the 22 general questions (95.45 % vs 81.82 %) is modest. A stronger control would retain the same node properties while ablating only the FDO packaging (or would evaluate against a conventional GraphRAG baseline that already stores metadata). Without that control the performance attribution remains confounded.
  2. [§IV.A] §IV.A.1–2: all quantitative results rest on the first 80 rows of a single GEO series and a hand-crafted set of 42 questions. While the authors correctly flag the absence of clinical validation, the limited scale also leaves open whether the reported accuracy and FAIR-compliance figures survive full-series construction, multi-document corpora, or physician-authored queries. At minimum the manuscript should quantify how many of the 42 questions can be answered from the raw tabular columns alone versus only after ontology enrichment.
minor comments (4)
  1. [§III.A] The formal model in §III.A introduces the set R of node types but never uses it in the subsequent definitions of ρ, λ or σ; a short clarifying sentence would help.
  2. [Table I] Table I lists R1.1 as “No” for both systems because the source dataset lacks a license; the text should state whether the authors attempted to attach a license during FDO creation or simply inherited the GEO restriction.
  3. [Fig. 6] Fig. 6 caption and surrounding text refer to “Fair Digital Object” with inconsistent capitalization; standardize to “FAIR Digital Object”.
  4. [References] The GitHub links [27],[28] are dated “accessed Sep. 19, 2025”; verify that the repositories remain publicly available and contain the exact scripts used for the reported tables.

Circularity Check

1 steps flagged

QA accuracy gains on metadata/ontology questions reduce largely by construction: Non-FAIR baseline omits exactly the PID/metadata/ontology properties those questions target.

specific steps
  1. other [Abstract; §IV.A.3 Baseline; §IV.B.2 / Tables II–III]
    "Beyond ensuring adherence to the FAIR principles, FAIR GraphRAG significantly improves question answering accuracy, coverage, and explainability, particularly for complex queries involving metadata and ontology links. … We introduce a Non-FAIR GraphRAG approach … FAIRification steps, such as assigning PIDs, metadata, and ontology terms, were skipped. … The Non-FAIR system fails to answer metadata and ontology-specific questions, resulting in 0% accuracy in these categories."

    The ablation that produces the Non-FAIR baseline removes precisely the node properties and ontology links that the 20 metadata/ontology questions ask for; the resulting 0 % accuracy (and the bulk of the overall 50-point gap) is therefore forced by the experimental construction rather than by any independent effect of FDO packaging or retrieval method. The claim that FAIRification itself yields the reported QA improvement is only partially isolated.

full rationale

The paper is an engineering/systems contribution, not a first-principles derivation, so classic mathematical circularity (self-definitional equations, uniqueness theorems, ansatz smuggling) is absent; no load-bearing self-citations or fitted parameters renamed as predictions appear. The sole partial circularity is empirical and confined to the central performance claim. The Non-FAIR baseline is produced by deliberately skipping the three FAIRification steps (PID assignment, metadata extraction, ontology-term mapping) that populate the very node properties and links queried by the 10 metadata-specific and 10 ontology-related questions. Consequently those categories score 0 % by design, inflating the headline overall gap (92.86 % vs 42.86 %). A residual independent gain remains on the 22 general questions (95.45 % vs 81.82 %), and the separate FAIR-indicator checklist (Table I) is non-circular. The evaluation therefore only weakly isolates any benefit of FDO packaging or GraphRAG retrieval from the mere presence of the additional fields. Score 4 reflects this partial, construction-driven component without over-penalizing an otherwise self-contained empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 1 invented entities

The paper is an empirical systems contribution; it rests on standard CS and FAIR assumptions rather than free parameters or invented physical entities. The main modeling choices are the property-graph formalization of FDOs and the decision to treat user-defined entity types as first-class nodes.

axioms (3)
  • domain assumption A property-graph model with node types {dataset, entity} and properties for PID, metadata and ontology links is a faithful and useful realization of FAIR Digital Objects.
    Stated in §III-A; underpins the entire formal model and construction pipeline.
  • domain assumption LLMs (gpt-4o-mini / Llama-3.3-70B) can reliably extract schema fields, metadata and Cypher queries when given few-shot prompts and a fixed JSON schema.
    Used throughout construction (§III-B) and retrieval (§III-C); performance differences between the two models are reported but not further analyzed.
  • ad hoc to paper Skipping only the PID/metadata/ontology steps produces a valid non-FAIR baseline for measuring the contribution of FAIRification.
    Defined in §IV-A3; the ablation design rests on this choice.
invented entities (1)
  • FAIR GraphRAG framework (FDO-as-node knowledge graph + LLM-driven construction and Cypher RAG) no independent evidence
    purpose: To make every retrieval unit FAIR by construction and thereby improve accuracy, coverage and explainability of domain QA.
    The central technical artifact introduced in the paper; no independent prior existence outside this work.

pith-pipeline@v1.1.0-grok45 · 17927 in / 2459 out tokens · 34169 ms · 2026-07-14T05:24:46.777160+00:00 · methodology

0 comments
read the original abstract

Retrieval-Augmented Generation (RAG) addresses the limitations of Large Language Models (LLMs) when providing responses to domain-specific questions. Graph-based RAG approaches, such as GraphRAG, enhance retrieval by capturing semantic relationships within knowledge graphs (KGs). While the FAIR principles (Findability, Accessibility, Interoperability, and Reusability) are becoming prevalent for scientific data management, especially in complex domains such as medicine, existing RAG approaches lack a structured FAIRification of the underlying knowledge resources. This lack limits their potential for FAIR information retrieval in these domains. To address this gap, we introduce FAIR GraphRAG, a novel framework that integrates FAIR Digital Objects (FDOs) as the fundamental units of a graph-based retrieval system. Each graph node represents an FDO that incorporates core data, metadata, persistent identifiers, and semantic links. We leverage LLMs to support schema construction and automated extraction of content and metadata from data sources. The framework was co-designed by physicians and computer scientists to ensure technical and clinical relevance. We apply FAIR GraphRAG to a biomedical dataset in gastroenterology, demonstrating its applicability to RNA-sequencing data. Beyond ensuring adherence to the FAIR principles, FAIR GraphRAG significantly improves question answering accuracy, coverage, and explainability, particularly for complex queries involving metadata and ontology links. This work shows the feasibility of combining FAIR data practices with graph-based retrieval techniques. We see potential for applying our approach to other specialized fields such as education and business.

Figures

Figures reproduced from arXiv: 2607.11464 by Carolin Victoria Schneider, Marlena Fl\"uh, Sandra Geisler, Soo-Yon Kim.

Figure 1
Figure 1. Figure 1: A High-level FAIR Digital Object model. Adopted from Wittenburg [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the FAIR GraphRAG framework. The underlying [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A FAIR knowledge graph representation with FDOs as nodes and [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Graph construction stages: Based on an input dataset, a FAIR [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Graph retrieval process: The prompt is prepared based on the user question and the knowledge graph schema. The graph retrieval engine prepares [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Chat Interface: The chat interface enables knowledge graph querying [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 1 canonical work pages

  1. [1]

    Large Language Models and Knowledge Graphs: Opportunities and Challenges,

    J. Z. Pan, S. Razniewski, J.-C. Kalo, S. Singhania, J. Chen, S. Dietze, H. Jabeen, J. Omeliyanenko, W. Zhang, M. Lissandrini, R. Biswas, G. de Melo, A. Bonifati, E. Vakaj, M. Dragoni, and D. Graux, “Large Language Models and Knowledge Graphs: Opportunities and Challenges,” Aug. 2023, arXiv:2308.06374 [cs]. [Online]. Available: http://arxiv.org/abs/2308.06374

    Pith/arXiv arXiv 2023

  2. [2]

    Graph retrieval-augmented generation: A survey,

    B. Peng, Y . Zhu, Y . Liu, X. Bo, H. Shi, C. Hong, Y . Zhang, and S. Tang, “Graph retrieval-augmented generation: A survey,” Sep. 2024, arXiv preprint arXiv:2408.08921. Available: arxiv.org/abs/2408.08921

    Pith/arXiv arXiv 2024

  3. [3]

    Graph rag in the wild: Insights and best practices from real- world applications,

    D. Collarana, C. I. Pack, Y .-Y . Liao, M. Fl ¨uh, J. Lehmkuhl, A. Nageri, A. Graß, M. Busch, P. Das, L. Dingels, S. Decker, and C. Beecks, “Graph rag in the wild: Insights and best practices from real- world applications,”Semantic Web Journal, 2025, under review. [Online]. Available: https://www.semantic-web-journal.net/content/ graph-rag-wild-insights-a...

    2025

  4. [4]

    Evaluation of Retrieval-Augmented Generation: A Survey,

    H. Yu, A. Gan, K. Zhang, S. Tong, Q. Liu, and Z. Liu, “Evaluation of Retrieval-Augmented Generation: A Survey,” inBig Data, W. Zhu, H. Xiong, X. Cheng, L. Cui, Z. Dou, J. Dong, S. Pang, L. Wang, L. Kong, and Z. Chen, Eds. Singapore: Springer Nature, 2025, pp. 102–120

    2025

  5. [5]

    Retrieval-Augmented Generation for AI-Generated Content: A Survey,

    P. Zhao, H. Zhang, Q. Yu, Z. Wang, Y . Geng, F. Fu, L. Yang, W. Zhang, J. Jiang, and B. Cui, “Retrieval-Augmented Generation for AI-Generated Content: A Survey,” Jun. 2024, arXiv:2402.19473 [cs]. [Online]. Available: http://arxiv.org/abs/2402.19473

    Pith/arXiv arXiv 2024

  6. [6]

    A Survey on Knowledge Graphs: Representation, Acquisition and Applications,

    S. Ji, S. Pan, E. Cambria, P. Marttinen, and P. S. Yu, “A Survey on Knowledge Graphs: Representation, Acquisition and Applications,” IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 2, pp. 494–514, Feb. 2022, arXiv:2002.00388 [cs]. [Online]. Available: http://arxiv.org/abs/2002.00388

    Pith/arXiv arXiv 2022

  7. [7]

    FAIR Digital Objects for Science: From Data Pieces to Actionable Knowledge Units,

    K. De Smedt, D. Koureas, and P. Wittenburg, “FAIR Digital Objects for Science: From Data Pieces to Actionable Knowledge Units,” Publications, vol. 8, no. 2, p. 21, Jun. 2020, number: 2 Publisher: Multidisciplinary Digital Publishing Institute. [Online]. Available: https://www.mdpi.com/2304-6775/8/2/21

    2020

  8. [8]

    Towards FAIRer Biological Knowledge Networks Using a Hybrid Linked Data and Graph Database Approach,

    M. Brandizi, A. Singh, C. Rawlings, and K. Hassani-Pak, “Towards FAIRer Biological Knowledge Networks Using a Hybrid Linked Data and Graph Database Approach,”Journal of Integrative Bioinformatics, vol. 15, no. 3, p. 20180023, Aug. 2018

    2018

  9. [9]

    CovidPubGraph: A FAIR Knowledge Graph of COVID-19 Publications,

    S. Pestryakova, D. V ollmers, M. A. Sherif, S. Heindorf, M. Saleem, D. Moussallem, and A.-C. N. Ngomo, “CovidPubGraph: A FAIR Knowledge Graph of COVID-19 Publications,”Scientific Data, vol. 9, no. 1, p. 389, Jul. 2022, publisher: Nature Publishing Group. [Online]. Available: https://www.nature.com/articles/s41597-022-01298-2

    2022

  10. [10]

    INGRIDKG: A FAIR Knowledge Graph of Graffiti,

    M. A. Sherif, A. A. M. da Silva, S. Pestryakova, A. F. Ahmed, S. Niemann, and A.-C. N. Ngomo, “INGRIDKG: A FAIR Knowledge Graph of Graffiti,”Scientific Data, vol. 10, no. 1, p. 318, May 2023

    2023

  11. [11]

    FAIR knowledge graphs with semantic units: A pro- totype,

    L. V ogt, “FAIR knowledge graphs with semantic units: A pro- totype,” Nov. 2023, arXiv preprint arXiv:2311.04761. Available: arxiv.org/abs/2311.04761

    Pith/arXiv arXiv 2023

  12. [12]

    Digital objects as drivers towards convergence in data infrastructures,

    P. Wittenburg, “Digital objects as drivers towards convergence in data infrastructures,” 2019, available: b2share.eudat.eu/records/

    2019

  13. [13]

    The FAIR Guiding Principles for scientific data management and stewardship,

    M. D. Wilkinson, M. Dumontier, I. J. Aalbersberg, G. Appleton, M. Axton, A. Baak, N. Blomberg, J.-W. Boiten, L. B. da Silva Santos, P. E. Bourne, J. Bouwman, A. J. Brookes, T. Clark, M. Crosas, I. Dillo, O. Dumon, S. Edmunds, C. T. Evelo, R. Finkers, A. Gonzalez-Beltran, A. J. G. Gray, P. Groth, C. Goble, J. S. Grethe, J. Heringa, P. A. C. ’t Hoen, R. Hoo...

    2016

  14. [14]

    Digital objects – fair digital objects: Which services are required?

    U. Schwardmann, “Digital objects – fair digital objects: Which services are required?”Data Science Journal, vol. 19, no. 1, Apr. 2020, available: datascience.codata.org/articles/10.5334/dsj-2020-015

    work page doi:10.5334/dsj-2020-015 2020

  15. [15]

    Applying the FAIR principles to data in a hospital: challenges and opportunities in a pandemic,

    N. Queralt-Rosinach, R. Kaliyaperumal, C. H. Bernab ´e, Q. Long, S. A. Joosten, H. J. van der Wijk, E. L. A. Flikkenschild, K. Burger, A. Jacobsen, B. Mons, M. Roos, BEAT-COVID Group, and COVID-19 LUMC Group, “Applying the FAIR principles to data in a hospital: challenges and opportunities in a pandemic,”Journal of Biomedical Semantics, vol. 13, no. 1, p....

    2022

  16. [16]

    Towards machine- actionable scientific knowledge as FAIR Digital Objects,

    L. V ogt, C. Biniossek, D. Betz, and M. Stocker, “Towards machine- actionable scientific knowledge as FAIR Digital Objects,” Mar. 2024

    2024

  17. [17]

    Linked Metadata for FAIR Digital Objects Carrying Computable Knowledge,

    A. Flynn, M. Conte, P. Boisvert, R. Richesson, Z. Landis-Lewis, and C. Friedman, “Linked Metadata for FAIR Digital Objects Carrying Computable Knowledge,”Research Ideas and Outcomes, vol. 8, p. e94438, Oct. 2022, publisher: Pensoft Publishers. [Online]. Available: https://riojournal.com/article/94438/

    2022

  18. [18]

    5.2 Quality-aware Knowledge Graph Construction,

    C. Cappiello, M. E. Vidal, S. Geisler, A. Iglesias-Molina, D. Van Assche, D. Chaves-Fraga, A. Dimou, I. Celino, A. Rula, and M. Lenzerini, “5.2 Quality-aware Knowledge Graph Construction,”Are Knowledge Graphs Ready for the Real World? Challenges and Perspective, vol. 20, p. 47, 2024. [Online]. Available: https://lirias.kuleuven.be/ retrieve/777295#page=49

    2024

  19. [19]

    Semantic units: organizing knowledge graphs into semantically meaningful units of representation,

    L. V ogt, T. Kuhn, and R. Hoehndorf, “Semantic units: organizing knowledge graphs into semantically meaningful units of representation,” Journal of Biomedical Semantics, vol. 15, no. 1, p. 7, May 2024. [Online]. Available: https://doi.org/10.1186/s13326-024-00310-5

    doi:10.1186/s13326-024-00310-5 2024

  20. [20]

    Docs2KG: Unified knowledge graph construction from het- erogeneous documents assisted by large language models,

    Q. Sun, Y . Luo, W. Zhang, S. Li, J. Li, K. Niu, X. Kong, and W. Liu, “Docs2KG: Unified knowledge graph construction from het- erogeneous documents assisted by large language models,” Jun. 2024, arXiv preprint arXiv:2406.02962. Available: arxiv.org/abs/2406.02962. Project: docs2kg.ai4wa.com

    Pith/arXiv arXiv 2024

  21. [21]

    iText2KG: Incremental Knowledge Graphs Construction Using Large Language Models,

    Y . Lairgi, L. Moncla, R. Cazabet, K. Benabdeslem, and P. Cl ´eau, “iText2KG: Incremental Knowledge Graphs Construction Using Large Language Models,” Sep. 2024, arXiv:2409.03284 [cs]. [Online]. Available: http://arxiv.org/abs/2409.03284

    Pith/arXiv arXiv 2024

  22. [22]

    Extract, Define, Canonicalize: An LLM- based Framework for Knowledge Graph Construction,

    B. Zhang and H. Soh, “Extract, Define, Canonicalize: An LLM- based Framework for Knowledge Graph Construction,” Oct. 2024, arXiv:2404.03868 [cs]. [Online]. Available: http://arxiv.org/abs/2404. 03868

    Pith/arXiv arXiv 2024

  23. [23]

    From Local to Global: A Graph RAG Approach to Query-Focused Summarization,

    D. Edge, H. Trinh, N. Cheng, J. Bradley, A. Chao, A. Mody, S. Truitt, and J. Larson, “From Local to Global: A Graph RAG Approach to Query-Focused Summarization,” Apr. 2024, arXiv:2404.16130. [Online]. Available: http://arxiv.org/abs/2404.16130

    Pith/arXiv arXiv 2024

  24. [24]

    B. T. Arnold, J. Theissen-Lipp, D. Collarana, C. Lange-Bever, S. Geisler, E. Curry, and S. J. Decker,Towards Enabling FAIR Dataspaces Using Large Language Models. CEUR-WS, 2024, medium: online Meeting Name: 21. Extended Semantic Web Conference

    2024

  25. [25]

    The Property Graph Database Model,

    R. Angles, “The Property Graph Database Model,” inProceedings of the 12th Alberto Mendelzon International Workshop on Foundations of Data Management, Cali, Colombia, May 21-25, 2018, ser. CEUR Workshop Proceedings, D. Olteanu and B. Poblete, Eds., vol. 2100. CEUR-WS.org, 2018. [Online]. Available: https://ceur-ws.org/V ol-2100/ paper26.pdf

    2018

  26. [26]

    Screening and molecular mechanism research on bile microRNAs associated with chemotherapy efficacy in perihilar cholangiocarcinoma,

    S. Fu, H. Du, Y . Dai, K. Zheng, G. Cao, L. Xu, Y . Zhong, C. Niu, Y . Kong, and X. Wang, “Screening and molecular mechanism research on bile microRNAs associated with chemotherapy efficacy in perihilar cholangiocarcinoma,”iScience, vol. 27, no. 12, p. 111437, Dec. 2024. [Online]. Available: https://www.sciencedirect.com/science/ article/pii/S2589004224026622

    2024

  27. [27]

    FAIR Knowledge Graph Construction,

    M. Fl ¨uh, “FAIR Knowledge Graph Construction,” https://github.com/ FAIR-GraphRAG/CSV2FAIR KG, 2025, accessed: Sep. 19, 2025

    2025

  28. [28]

    FAIR Retrieval-Augmented Generation Interface,

    ——, “FAIR Retrieval-Augmented Generation Interface,” https://github. com/FAIR-GraphRAG/RAG-interface, Sep. 2025, accessed: Sep. 19, 2025

    2025