Pith. sign in

REVIEW 5 major objections 5 minor 5 cited by

Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A system that retrieves relevant subgraphs from Wikidata and inserts them into LLM prompts improves schema-matching precision and F1 over LLM and PLM baselines, with no retraining of the backbone model.

desk verdict KG-RAG4SM's idea is fresh but the evaluation's baseline operating points are degenerate, so the SOTA claims don't hold — still worth reviewing for the dataset and mechanism. read the letter →

arxiv 2501.08686 v1 pith:LQY2CCXP submitted 2025-01-15 cs.DB cs.CLcs.IR

Chuangtao Ma , Sriom Chakrabarti , Arijit Khan , Bálint Molnár This is my paper · ORCID
classification cs.DBcs.CLcs.IR
keywords schemamatchingknowledge-graphretrievalretrieval-augmentedgenerationlargelanguagemodelsdataintegrationWikidatahealthcaresubgraphranking
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

KG-RAG4SM is a retrieval-augmented generation pipeline for schema matching: given a pair of attributes from a source and a target database, it retrieves the most relevant triples and paths from Wikidata, ranks them, and pastes the chosen subgraph into the prompt of an LLM that answers a binary question about whether the two attributes are semantically matched. The paper's central claim is that this external knowledge injection resolves semantic ambiguities that similarity-based and plain-LLM matchers miss, and that it does so without fine-tuning or training. On healthcare benchmarks, the reported effect is sizable: with Jellyfish-8B as the backbone, precision and F1 on MIMIC improve by 35.89% and 30.50% over the same model without retrieval; with GPT-4o-mini as the backbone, precision and F1 on Synthea improve by 69.20% and 21.97% over the trained SMAT baseline. The paper also introduces EMED, a new real-world healthcare schema-matching benchmark, and argues that the retrieval is efficient enough to scale to Wikidata-scale graphs.

What carries the argument

KG-RAG4SM (knowledge graph-based retrieval-augmented generation for schema matching) is a pipeline that combines retrieval, ranking, and prompt augmentation. It forms a binary question from the source attribute, target attribute, and their textual descriptions; retrieves relevant knowledge from Wikidata; and then ranks the retrieved paths before injecting the top-1 or top-2 subgraphs into the LLM prompt. Two retrieval routes carry the argument: vector-based KG-triple retrieval, which embeds questions and KG triples and selects by cosine similarity, and vector-based entity retrieval followed by 3-hop BFS traversal between retrieved entity pairs. A ranking scheme scores paths by how many of the top retrieved relations they contain, normalized by path length, so that concise, relevant subgraphs are preferred over long or noisy ones. The load-bearing step is the subgraph itself: it gives the LLM a verifiable external semantic relationship instead of relying only on parametric memory.

What would settle it

Hold out the exact prompt-example pairs from the CMS ground truth, rerun the pipeline, and compare F1; if the gain over the no-RAG baseline concentrates on those pairs and collapses on the remaining ones, the reported improvement is prompt memorization rather than retrieval benefit.

Watch

Extended reading notes

Core claim

KG-RAG4SM treats each candidate attribute pair as a binary schema-matching question and augments the prompt with a small, ranked subgraph retrieved from Wikidata. The central claim is that the retrieved subgraph supplies the missing commonsense or domain knowledge: for example, knowing that a beneficiary is a subclass of a customer and a customer is a subclass of a patient helps the model separate measurement time from intervention date, while knowing that a physical-status classification is a medical classification helps it align two identifier columns. Across MIMIC, Synthea, CMS, and the new EMED dataset, the method reports higher precision and F1 than Jellyfish-8B, GPT-4o-mini without retrieval, SMAT, and Unicorn, with the two vector-based retrievers (KG-triple retrieval and entity retrieval followed by 3-hop BFS) as the best-performing pipelines. The paper further claims that the retrieved context mitigates LLM hallucination on complex real-world mapping cases.

Load-bearing premise

The load-bearing premise is that the few-shot examples in the prompt are disjoint from the attribute pairs being evaluated; the paper does not enforce a train/test split, and the illustrative example pairs come from the CMS benchmark, which is itself one of the evaluated datasets.

Editorial extensions

If this is right

  • Schema matching can be improved without a label-hungry training phase: KG-RAG4SM needs only prompt examples and an external KG, not a supervised training set.
  • Smaller LLMs can outperform larger ones when equipped with retrieved KG context, because retrieved knowledge can conflict with the parametric knowledge of large models.
  • Noise control is central: keeping only top-1 or top-2 ranked subgraphs prevents context poisoning, and on the CMS dataset ranking-based refinement improved F1 by 62.59% over unranked retrieval for the KG-triples pipeline.
  • The paradigm is designed to transfer to other LLM-based data integration tasks such as entity matching and data fusion, which the paper names as future work.
  • A commonsense KG such as Wikidata can support domain-specific healthcare matching for ambiguities involving acronyms, abbreviations, relatedness, and hierarchical roles.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not establish a clean separation between prompt examples and evaluated pairs, so the CMS results are the least protected against memorization; MIMIC and Synthea numbers are more informative as tests of the retrieval claim.
  • The reported improvements are relative gains over baselines whose absolute precision and F1 are low, so the practical significance depends on whether a deployment can tolerate many false positives or would need a human-in-the-loop.
  • A stronger test of the mechanism would compare KG-RAG4SM against a retrieval-augmented baseline that retrieves from the schema's own textual descriptions rather than from Wikidata, isolating the value of external structured knowledge from the value of any retrieved context.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes KG-RAG4SM, a knowledge-graph-based retrieval-augmented generation framework for schema matching. It retrieves relevant subgraphs from Wikidata using vector-based, BFS-based, LLM-based, and query-based methods, prunes them with ranking schemes, and injects the retrieved knowledge into LLM prompts. The authors introduce a new EMED benchmark, report experiments on MIMIC, Synthea, CMS, and EMED, and claim that KG-RAG4SM outperforms state-of-the-art LLM-based (Jellyfish-8B, GPT-4o-mini) and PLM-based (SMAT, Unicorn) matchers in precision and F1, while also mitigating LLM hallucinations. The paper includes ablation studies on retrieval strategies, backbone LLMs, and ranking, plus efficiency comparisons.

Significance. If the stated results held, the paper would offer a practical way to inject structured world knowledge into LLM-based schema matching without retraining, and the EMED dataset would be a useful resource for the community. The paper ships code and data, and the ablation of four retrieval paradigms is informative. The case studies give concrete illustrations of how KG context can correct LLM errors. However, the headline claims are not currently supported by the evidence: the comparison baselines are operated at degenerate working points, the configuration is selected using test data, the few-shot examples overlap with the evaluated benchmark, and the absolute precision values are too low for real deployment. These issues materially weaken the significance of the empirical claims.

major comments (5)
  1. [§4.2.1, Tables 5 and 6] The SOTA comparison rests on degenerate baseline operating points. Jellyfish-8B reaches 100% recall with precision 6.81 on MIMIC, SMAT reaches 100% recall with precision 11.82 on Synthea, and Unicorn-trained reaches roughly 99.99% recall with precision below 1% on all datasets. The paper does not tune decision thresholds for these baselines; it simply compares F1 and precision at these extreme points. Under the heavy class imbalance of these benchmarks (e.g., 129 positive pairs out of 64,080 on MIMIC), a model that predicts almost all pairs as positive trivially obtains high recall, and the reported relative gains over such baselines do not constitute evidence of SOTA-level schema matching. The evaluation should either tune each baseline's threshold on a validation set or report precision-recall curves.
  2. [§4.1.1, §3.2, §4.3.1] The evaluation protocol has no train/validation/test split for KG-RAG4SM, and the system configuration is selected after observing CMS test results. The paper states that cosine similarity was chosen 'after extensive experimentation' (§3.2.1), that top-2 ranked subgraphs are 'generally more effective' (§3.2.3), and that 3-hop search is a 'trade-off solution' (§3.2.2); all of these decisions are validated on the CMS dataset (§4.3.1). Reporting results on the same dataset used to select the pipeline, top-k, BFS depth, and similarity metric inflates the reported performance and makes the transfer to other datasets difficult to assess. The authors should hold out a validation set for configuration selection or report results under a predefined configuration.
  3. [§3.1, Figure 4, §4.1.1] The few-shot examples in the prompts are not guaranteed to be disjoint from the evaluation pairs. Figure 4 and Example 1 of Figure 8 use the pair death-personid and beneficiarysummary-desynpufid, which are CMS attributes, and CMS is one of the evaluated datasets. The paper says there is no train/validation/test split for KG-RAG4SM (§4.1.1) but does not state that these prompt examples are excluded from the evaluation. If any of the few-shot pairs coincide with an evaluated pair, the reported gains include answer memorization rather than retrieval benefit. The authors must explicitly confirm the disjointness of prompt examples and test pairs, or re-design the evaluation with examples from an unrelated domain.
  4. [Abstract and §4.2.1, Table 5] The headline improvement numbers in the abstract are not reproducible from the reported tables. The abstract claims a 35.89% precision improvement on MIMIC over Jellyfish-8B, but Table 5 shows 9.52% versus 6.81%, which is a 39.8% relative increase. The F1 improvement (30.50%) is close to the computed 30.56% (16.66 versus 12.76), but the precision figure is inconsistent. This discrepancy suggests that the abstract numbers are not tightly coupled to the experimental results and must be corrected.
  5. [§4.2, Table 4] The experiments report a single run for each LLM-based configuration, with no error bars or repeated runs. Although the decoding hyperparameters (top-k=1, top-p=0.9, temperature 0.6) likely make the outputs nearly deterministic, the paper does not establish this or report any variance. Given the small metric differences among some settings (e.g., CMS precision 52.38 versus 44.44 in Table 8), a single run is insufficient to demonstrate that the differences are stable. The paper should report multiple runs and variance, or explicitly justify that the decoding settings remove sampling stochasticity.
minor comments (5)
  1. [§4.2.1] The text contains a typo: 'MIMC' should be 'MIMIC' in the results analysis.
  2. [Figure 4] The figure title reads 'Prompts for LLMs as Schema Mather' and the prompt text contains 'atttribute'; both are typos.
  3. [§6] The conclusion refers to the method as 'KG-RAG4M' instead of 'KG-RAG4SM'.
  4. [§2.1.2 and §2.2] The symbol G is used both for the large knowledge graph and for the retrieved subgraph, which is confusing; the paper should use distinct notation for the full graph and the subgraph.
  5. [Table 4] The hyperparameter table lists top-k=1 for LLM generation; with top-k=1 the temperature 0.6 has no effect on sampling. The authors should clarify whether sampling is enabled and how the selection of top-k=1 interacts with the temperature setting.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are empirical outcomes of an independently specified retrieval pipeline, not reductions of the method to its inputs.

full rationale

The paper's central claim is an empirical comparison: KG-RAG4SM retrieves Wikidata subgraphs by vector similarity, BFS traversal, or query generation, injects them into LLM prompts, and reports precision/F1 on four benchmarks. There is no derivation chain in which a predicted quantity is defined as, or algebraically identical to, a fitted input. The retrieval modules (cosine similarity over RoBERTa embeddings, HNSW search, 3-hop BFS, frequency/length-normalized ranking) are specified independently of the matching labels. The statement in §4.1.1 that 'there is no splitting of training, validation, and test datasets for our KG-RAG4SM technique' and the use of few-shot examples drawn from the same benchmark family are genuine evaluation-design weaknesses: they risk answer memorization for the handful of example pairs and allow pipeline selection on test data. However, these issues do not make the reported F1/P values equal to the prompt contents by construction; the overwhelming majority of test pairs are not shown as examples, and the baselines' degenerate operating points (e.g., 100% recall with 6.81% precision for Jellyfish on MIMIC in Table 5) are a baseline-configuration concern, not a circularity. No load-bearing self-citation or imported uniqueness theorem appears: self-citations [3], [13], [40] concern the e-MedSolution case study and are not used to justify the retrieval design. Accordingly, the appropriate finding is no significant circularity, with any concerns about data leakage and benchmark validity assigned to experimental rigor rather than circularity.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

No new physical or mathematical entities are introduced. The ledger captures the system configuration and dataset assumptions that the empirical claims depend on.

free parameters (5)
  • Top-k retrieved subgraphs = 2
    Top-2 chosen for main results after experiments on CMS (§4.3.1, Table 9).
  • BFS hop depth = 3
    Dmax=3 picked as an empirical trade-off (§3.2.2).
  • Subgraph retrieval pipeline = Vector-based KG triples retrieval
    Selected as best pipeline after CMS experiments (§3.3, §4.3.1).
  • Similarity metric = Cosine similarity
    Chosen after experimentation against Euclidean and Manhattan distances (§3.2.1).
  • Ranking normalization = unspecified
    Frequency score normalized by path length; exact weighting not given (§3.2.3).
assumptions (6)
  • domain assumption Ground-truth labels in OMAP and EMED are correct and complete.
    All metrics are computed against these labels (§4.1.1); no label-quality or inter-annotator analysis is provided.
  • domain assumption Wikidata contains relevant subgraphs for arbitrary schema attribute pairs.
    The method depends on KG coverage for each question (§3.2); no coverage statistics are reported.
  • domain assumption LLMs can correctly parse semi-structured subgraph text in prompts.
    Triples with arrows are inserted directly into prompts (§3.2.4); the paper does not verify parsing robustness.
  • standard math Sentence embeddings of questions and KG triples preserve semantic relatedness.
    RoBERTa/ChromaDB retrieval assumes vector proximity correlates with matching relevance (§3.2.1).
  • ad hoc to paper Few-shot prompt examples are disjoint from evaluation pairs.
    No train/validation/test split is made for KG-RAG4SM (§4.1.1), and examples in Figures 4 and 8 resemble CMS benchmark pairs.
  • domain assumption One generation per question with temperature 0.6 is a stable estimator.
    No repeated sampling is reported, so the paper implicitly assumes single-run results are representative (Table 4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching." pith.science (2026). https://pith.science/paper/LQY2CCXP

@misc{pith2026250108686,
  author       = {Pith},
  title        = {Pith review of: Knowledge Graph-based Retrieval-Augmented Generation for Schema Matching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LQY2CCXP}},
  note         = {Machine review of arXiv:2501.08686}
}
read the original abstract

Traditional similarity-based schema matching methods are incapable of resolving semantic ambiguities and conflicts in domain-specific complex mapping scenarios due to missing commonsense and domain-specific knowledge. The hallucination problem of large language models (LLMs) also makes it challenging for LLM-based schema matching to address the above issues. Therefore, we propose a Knowledge Graph-based Retrieval-Augmented Generation model for Schema Matching, referred to as the KG-RAG4SM. In particular, KG-RAG4SM introduces novel vector-based, graph traversal-based, and query-based graph retrievals, as well as a hybrid approach and ranking schemes that identify the most relevant subgraphs from external large knowledge graphs (KGs). We showcase that KG-based retrieval-augmented LLMs are capable of generating more accurate results for complex matching cases without any re-training. Our experimental results show that KG-RAG4SM outperforms the LLM-based state-of-the-art (SOTA) methods (e.g., Jellyfish-8B) by 35.89% and 30.50% in terms of precision and F1 score on the MIMIC dataset, respectively; KG-RAG4SM with GPT-4o-mini outperforms the pre-trained language model (PLM)-based SOTA methods (e.g., SMAT) by 69.20% and 21.97% in terms of precision and F1 score on the Synthea dataset, respectively. The results also demonstrate that our approach is more efficient in end-to-end schema matching, and scales to retrieve from large KGs. Our case studies on the dataset from the real-world schema matching scenario exhibit that the hallucination problem of LLMs for schema matching is well mitigated by our solution.

Figures

Figures reproduced from arXiv: 2501.08686 by the authors.

Figure 1
Figure 1. Example of schema matching in the EHR data model. Table (a) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The Role of KG context in augmenting LLMs for schema match [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of our proposed KG-RAG4SM method. Given schema matching questions and an external knowledge graph, we (1) retrieve the relevant KG triplets based on vector similarity between questions embeddings and KG triplet embeddings; (2) prune the retrieved relevant KG triplets with vector similarity-based ranking; (3) augment prompts with the retrieved and refined subgraphs from large-size KG and gener￾ate the final … view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Prompts for LLMs as schema matcher. 3.2.1 Retrieving Relevant Entities from the KG Before retrieving the subgraph from KG, we need to retrieve the entities from large-size KGs that are relevant to the given question of schema matching. We follow one of the two develope…
Figure 5
Figure 5. Figure 5: The placeholders {question} are replaced with the real questions from our selected dataset. Nevertheless, LLMs can hallucinate, thus it may report some entities that do not, in fact, exist in the commonsense KG, thus, we need to verify the retrieved entities by LLMs be…
Figure 5
Figure 5. Figure 5: Prompts for LLMs as entity retriever. it well-suited for applications in schema matching and se￾mantic search. Offline KG embedding and vector storage. The embeddings for KG entities and relations are generated in an offline manner using pre-trained transformer-based l…
Figure 6
Figure 6. Figure 6: Prompts for LLMs as subgraph retriever. Prompts for LLMs to Cypher Query for Subgraph Retrieval You are a neo4j graph database expert, I will give the following file (relation￾ship type, node label, property key) including the schema information of the wikidata neo4j g…
Figure 7
Figure 7. Figure 7: Prompts for LLMs to generate Cypher query for retrieval. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Prompts for augmenting generation of LLMs with the retrieved [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Average number of retrieved subgraphs for different subgraph [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Case study: Hallucinations mitigation of [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. KAMR: Grounding Generation via Knowledge-Aligned Multi-hop Retrieval

    cs.IR 2026-07 conditional novelty 6.0 of 10

    Partial-alignment contrastive pretraining plus anchor-then-expand graph retrieval improves multi-hop KG evidence recovery and downstream QA over strong dense and graph RAG baselines.

  2. A Survey of LLM $\times$ DATA

    cs.DB 2025-05 conditional novelty 5.0 of 10

    A comprehensive survey of the bidirectional links between LLMs and data management, organized as DATA4LLM and LLM4DATA with a new 'IaaS' data-quality framework.

  3. Towards Trustworthy and Cost-Efficient Data Integration: From Na\"ive RAG to Agentic RAG

    cs.DB 2026-07 conditional novelty 4.0 of 10

    The paper argues that agentic RAG with adaptive retrieval, iterative reasoning, and graph memory is the path to trustworthy and cost-efficient data integration, and sketches a six-agent architecture.

  4. Learning-Infused Formal Reasoning: From Contract Synthesis to Artifact Reuse and Formal Semantics

    cs.SE 2026-02 unverdicted novelty 4.0 of 10

    A vision paper proposing Learning-Infused Formal Reasoning (LIFR), a hybrid LLM+graph framework for contract synthesis, artifact reuse, and semantic foundations in verification.

  5. A Comprehensive Survey on Integrating Large Language Models with Knowledge-Based Methods

    cs.CL 2025-01 conditional novelty 3.0 of 10

    A narrative review of LLM knowledge integration that categorizes techniques and compiles benchmarks, but lacks a systematic method and contains unreliable citations.

Reference graph

Works this paper leans on

60 extracted references · 49 canonical work pages · cited by 5 Pith papers

  1. [1]

    D. H. Hai, Schema Matching and Mapping-Based Data Integration: Architecture, Approaches and Evaluation. VDM Verlag, 2007

    work page 2007

  2. [2]

    Rahm, Towards Large-Scale Schema and Ontology Matching

    E. Rahm, Towards Large-Scale Schema and Ontology Matching . Springer, 2011, pp. 3–27

    work page 2011

  3. [3]

    Knowledge-enriched Schema Mapping: A Preliminary Case Study of e-MedSolution System,

    C. Ma, “Knowledge-enriched Schema Mapping: A Preliminary Case Study of e-MedSolution System,” E¨ otv¨ os Lor´ and University, Tech. Rep., 2020

    work page 2020

  4. [4]

    A Consensus-based Approach for Harmonizing the OHDSI Common Data Model with HL7 FHIR,

    G. Jiang, R. C. Kiefer, D. K. Sharma, E. Prud’hommeaux, and H. R. Solbrig, “A Consensus-based Approach for Harmonizing the OHDSI Common Data Model with HL7 FHIR,” Studies in Health Technology and Informatics, vol. 245, p. 887, 2017

    work page 2017

  5. [5]

    Schema Matching with Large Language Models: an Experimental Study,

    M. Parciak, B. Vandevoort, F. Neven, L. M. Peeters, and S. Van- summeren, “Schema Matching with Large Language Models: an Experimental Study,” in TaDA: International Workshop on Tabular Data Analysis, 2024

    work page 2024

  6. [6]

    Jellyfish: Instruction-Tuning Local Large Language Models for Data Pre- processing,

    H. Zhang, Y. Dong, C. Xiao, and M. Oyamada, “Jellyfish: Instruction-Tuning Local Large Language Models for Data Pre- processing,” in Empirical Methods in Natural Language Processing (EMNLP), 2024, pp. 8754–8782

    work page 2024

  7. [7]

    Fine-tuning Large Language Models for Entity Matching,

    A. Steiner, R. Peeters, and C. Bizer, “Fine-tuning Large Language Models for Entity Matching,”arXiv preprint arXiv:2409.08185, 2024

    arXiv 2024

  8. [8]

    Cost-Aware Uncertainty Re- duction in Schema Matching with GPT-4: The Prompt-Matcher Framework,

    L. Feng, H. Li, and C. J. Zhang, “Cost-Aware Uncertainty Re- duction in Schema Matching with GPT-4: The Prompt-Matcher Framework,” arXiv preprint arXiv:2408.14507, 2024

    arXiv 2024

Show all 60 references
  1. [9]

    ReMatch: Re- trieval Enhanced Schema Matching with LLMs,

    E. Sheetrit, M. Brief, M. Mishaeli, and O. Elisha, “ReMatch: Re- trieval Enhanced Schema Matching with LLMs,” arXiv preprint arXiv:2403.01567, 2024

    2024 arXiv

  2. [10]

    Schema Matching using Pre-trained Language Models,

    Y. Zhang, A. Floratou, J. Cahoon, S. Krishnan, A. C. M ¨uller, D. Banda, F. Psallidas, and J. M. Patel, “Schema Matching using Pre-trained Language Models,” in International Conference on Data Engineering (ICDE), 2023, pp. 1558–1571

    2023

  3. [11]

    Uni- corn: A Unified Multi-tasking Model for Supporting Matching Tasks in Data Integration,

    J. Tu, J. Fan, N. Tang, P . Wang, G. Li, X. Du, X. Jia, and S. Gao, “Uni- corn: A Unified Multi-tasking Model for Supporting Matching Tasks in Data Integration,” Proceedings of the ACM on Management of Data, vol. 1, no. 1, pp. 1–26, 2023

    2023

  4. [12]

    Using Pre-trained Language Models to Resolve Textual and Semantic Merge Conflicts (Experience Paper),

    J. Zhang, T. Mytkowicz, M. Kaufman, R. Piskac, and S. K. Lahiri, “Using Pre-trained Language Models to Resolve Textual and Semantic Merge Conflicts (Experience Paper),” in ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) , 2022, pp. 77–88

    2022

  5. [13]

    Knowledge Enriched Schema Matching Framework for Heterogeneous Data Integra- tion,

    C. Ma, B. Moln ´ar, ´A. Tarcsi, and A. Bencz´ur, “Knowledge Enriched Schema Matching Framework for Heterogeneous Data Integra- tion,” in Information Technology and Data Science (CITDS), 2022, pp. 183–188

    2022

  6. [14]

    SMed- BERT: A Knowledge-Enhanced Pre-trained Language Model with Structured Semantics for Medical Text Mining,

    T. Zhang, Z. Cai, C. Wang, M. Qiu, B. Yang, and X. He, “SMed- BERT: A Knowledge-Enhanced Pre-trained Language Model with Structured Semantics for Medical Text Mining,” in Annual Meeting of the Association for Computational Linguistics and International Joint Conference on Natu...

    2021

  7. [15]

    Knowledge Enhanced Pretrained Language Models: A Compreshensive Sur- vey,

    X. Wei, S. Wang, D. Zhang, P . Bhatia, and A. Arnold, “Knowledge Enhanced Pretrained Language Models: A Compreshensive Sur- vey,” arXiv preprint arXiv:2110.08455, 2021

    2021 arXiv

  8. [16]

    Knowledge Graph En- hanced Language Models for Sentiment Analysis,

    J. Li, X. Li, L. Hu, Y. Zhang, and J. Wang, “Knowledge Graph En- hanced Language Models for Sentiment Analysis,” in International Semantic Web Conference (ISWC), 2023, pp. 447–464

    2023

  9. [17]

    REAL: A Retrieval-Augmented Entity Linking Approach for Biomedical Concept Recognition,

    D. Shlyk, T. Groza, M. Mesiti, S. Montanelli, and E. Cavalleri, “REAL: A Retrieval-Augmented Entity Linking Approach for Biomedical Concept Recognition,” in Workshop on Biomedical Nat- ural Language Processing, 2024, pp. 380–389

    2024

  10. [18]

    Wikidata: A Free Collaborative Knowledgebase,

    D. Vrande ˇci´c and M. Kr ¨otzsch, “Wikidata: A Free Collaborative Knowledgebase,” Communications of the ACM , vol. 57, no. 10, pp. 78–85, 2014

    2014

  11. [19]

    SNOMED CT Concept Hierarchies for Sharing Definitions of Clinical Conditions Using Electronic Health Record Data,

    D. L. Willett, V . Kannan, L. Chu, J. R. Buchanan, F. T. Velasco, J. D. Clark, J. S. Fish, A. R. Ortuzar, J. E. Youngblood, D. G. Bhat, and M. A. Basit, “SNOMED CT Concept Hierarchies for Sharing Definitions of Clinical Conditions Using Electronic Health Record Data,” Appl Cli...

    2018

  12. [20]

    RACOON: An LLM-based Framework for Retrieval-Augmented Column Type Annotation with a Knowledge Graph,

    L. Wei, G. Xiao, and M. Balazinska, “RACOON: An LLM-based Framework for Retrieval-Augmented Column Type Annotation with a Knowledge Graph,” arXiv preprint arXiv:2409.14556, 2024

    2024 arXiv

  13. [21]

    SMAT: An Attention- based Deep Learning Solution to the Automation of Schema Matching,

    J. Zhang, B. Shin, J. D. Choi, and J. C. Ho, “SMAT: An Attention- based Deep Learning Solution to the Automation of Schema Matching,” in European Conference on Advances in Databases and Information Systems (ADBIS), 2021, pp. 260–274. 17

    2021

  14. [22]

    GRAG: Graph Retrieval-Augmented Generation,

    Y. Hu, Z. Lei, Z. Zhang, B. Pan, C. Ling, and L. Zhao, “GRAG: Graph Retrieval-Augmented Generation,” arXiv preprint arXiv:2405.16506, 2024

    2024 arXiv

  15. [23]

    HybridRAG: Integrating Knowledge Graphs and Vector Re- trieval Augmented Generation for Efficient Information Extrac- tion,

    B. Sarmah, B. Hall, R. Rao, S. Patel, S. Pasquali, and D. Mehta, “HybridRAG: Integrating Knowledge Graphs and Vector Re- trieval Augmented Generation for Efficient Information Extrac- tion,” arXiv preprint arXiv:2408.04948, 2024

    2024 arXiv

  16. [24]

    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,” arXiv preprint arXiv:2408.08921, 2024

    2024 arXiv

  17. [25]

    Knowledge Graph- Enhanced Large Language Models via Path Selection,

    H. Liu, S. Wang, Y. Zhu, Y. Dong, and J. Li, “Knowledge Graph- Enhanced Large Language Models via Path Selection,” in Findings of the Association for Computational Linguistics (ACL 2024), 2024, pp. 6311–6321

    2024

  18. [26]

    RoBERTa: A Ro- bustly Optimized BERT Pretraining Approach,

    Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettlemoyer, and V . Stoyanov, “RoBERTa: A Ro- bustly Optimized BERT Pretraining Approach,” arXiv preprint arXiv:1907.11692, 2019

    1907 arXiv

  19. [27]

    Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks,

    N. Reimers and I. Gurevych, “Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks,” arXiv preprint arXiv:1908.10084, 2019

    1908 arXiv

  20. [28]

    Efficient and Robust Approxi- mate Nearest Neighbor Search using Hierarchical Navigable Small World Graphs,

    Y. A. Malkov and D. A. Yashunin, “Efficient and Robust Approxi- mate Nearest Neighbor Search using Hierarchical Navigable Small World Graphs,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 4, pp. 824–836, 2018

    2018

  21. [29]

    Retrieve- Rewrite-Answer: A KG-to-Text Enhanced LLMs Framework for Knowledge Graph Question Answering,

    Y. Wu, N. Hu, S. Bi, G. Qi, J. Ren, A. Xie, and W. Song, “Retrieve- Rewrite-Answer: A KG-to-Text Enhanced LLMs Framework for Knowledge Graph Question Answering,” in International Joint Conference on Knowledge Graphs (IJCKG), 2023

    2023

  22. [30]

    MIMIC-III, A Freely Accessible Critical Care Database,

    A. E. Johnson, T. J. Pollard, L. Shen, L. wei H. Lehman, M. Feng, M. Ghassemi, B. Moody, P . Szolovits, L. A. Celi, and R. G. Mark, “MIMIC-III, A Freely Accessible Critical Care Database,” Scientific Data, vol. 3, 2016

    2016

  23. [31]

    Synthea: An Approach, Method, and Software Mechanism for Generat- ing Synthetic Patients and the Synthetic Electronic Health Care Record,

    J. Walonoski, M. Kramer, J. Nichols, A. Quina, C. Moesel, D. Hall, C. Duffett, K. Dube, T. Gallagher, and S. McLachlan, “Synthea: An Approach, Method, and Software Mechanism for Generat- ing Synthetic Patients and the Synthetic Electronic Health Care Record,” Journal of the Am...

    2018

  24. [32]

    Transforming the 2.33 M- patient Medicare synthetic public use files to the OMOP CDMv5: ETL-CMS software and processed data available and feature- complete,

    C. G. Lambert and P . Amritansh Kumar, “Transforming the 2.33 M- patient Medicare synthetic public use files to the OMOP CDMv5: ETL-CMS software and processed data available and feature- complete,” https://github.com/OHDSI/ETL-CMS, 2016

    2016

  25. [33]

    OHDSI, 2019

    OHDSI, The Book of OHDSI: Observational Health Data Sciences and Informatics. OHDSI, 2019

    2019

  26. [34]

    An Analysis of Links in Wikidata,

    A. Haller, A. Polleres, D. Dobriy, N. Ferranti, and S. J. Rodr´ıguez M ´endez, “An Analysis of Links in Wikidata,” in The Semantic Web: International Conference (ESWC), 2022, p. 21–38

    2022

  27. [35]

    Attention- over-Attention Neural Networks for Reading Comprehension,

    Y. Cui, Z. Chen, S. Wei, S. Wang, T. Liu, and G. Hu, “Attention- over-Attention Neural Networks for Reading Comprehension,” in Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (ACL), 2017, pp. 593–602

    2017

  28. [36]

    Matchmaker: Self-Improving Large Language Model Programs for Schema Matching,

    N. Seedat and M. van der Schaar, “Matchmaker: Self-Improving Large Language Model Programs for Schema Matching,” arXiv preprint arXiv:2410.24105, 2024

    2024 arXiv

  29. [37]

    KcMF: A Knowledge- compliant Framework for Schema and Entity Matching with Fine- tuning-free LLMs,

    Y. Xu, H. Li, K. Chen, and L. Shou, “KcMF: A Knowledge- compliant Framework for Schema and Entity Matching with Fine- tuning-free LLMs,” arXiv preprint arXiv:2410.12480, 2024

    2024 arXiv

  30. [38]

    Deberta: decoding-Enhanced Bert with Disentangled Attention,

    P . He, X. Liu, J. Gao, and W. Chen, “Deberta: decoding-Enhanced Bert with Disentangled Attention,” in International Conference on Learning Representations (ICLR), 2021

    2021

  31. [39]

    Schema Mapping,

    A. Fuxman and R. J. Miller, “Schema Mapping,” in Encyclopedia of Database Systems. Springer, 2009, pp. 2481–2488

    2009

  32. [40]

    Knowledge Enhanced Schema Matching for Heteroge- neous Data Integration,

    C. Ma, “Knowledge Enhanced Schema Matching for Heteroge- neous Data Integration,” Ph.D. dissertation, E ¨otv¨os Lor ´and Uni- versity, Budapest, Hungary, 2024

    2024

  33. [41]

    Database Schema Matching using Ma- chine Learning with Feature Selection,

    J. Berlin and A. Motro, “Database Schema Matching using Ma- chine Learning with Feature Selection,” in Advanced Information Systems Engineering (CAiSE), 2002, pp. 452–466

    2002

  34. [42]

    Combining Lexical and Semantic Similarity Mea- sures with Machine Learning Approach for Ontology and Schema Matching Problem,

    L. Bulygin, “Combining Lexical and Semantic Similarity Mea- sures with Machine Learning Approach for Ontology and Schema Matching Problem,” in Data Analytics and Management in Data Intensive Domains ( DAMDID/RCDL), 2018, pp. 245–249

    2018

  35. [43]

    Quickmig: Automatic Schema Matching for Data Migration Projects,

    C. Drumm, M. Schmitt, H.-H. Do, and E. Rahm, “Quickmig: Automatic Schema Matching for Data Migration Projects,” inACM conference on Conference on Information and Knowledge Management (CIKM), 2007, pp. 107–116

    2007

  36. [44]

    Schema Matching Based on Attribute Values and Background Ontology,

    A. Nathalie, “Schema Matching Based on Attribute Values and Background Ontology,” in AGILE International Conference on Geo- graphic Information Science, 2009, pp. 1–9

    2009

  37. [45]

    Medical Schema Matching using Knowledge Graph Embedding,

    C. Gao, T. Wu, S. Jing, and Y. Wang, “Medical Schema Matching using Knowledge Graph Embedding,” in Workshop on Extraction and Evaluation of Knowledge Entities from Scientific Documents , ser. CEUR Workshop Proceedings, vol. 3210, 2022, pp. 64–68

    2022

  38. [46]

    A Knowledge Graph Enhanced Semantic Matching Method for Plan Recommendation,

    R. Liang, S. Zheng, K. Deng, Z. Mao, W. Ma, and Z. Zhang, “A Knowledge Graph Enhanced Semantic Matching Method for Plan Recommendation,” in International Conference on Swarm Intelligence (ICSI), 2021, pp. 290–299

    2021

  39. [47]

    Data Integration and Machine Learn- ing: A Natural Synergy,

    L. Dong and T. Rekatsinas, “Data Integration and Machine Learn- ing: A Natural Synergy,” Proc. VLDB Endow. , vol. 11, no. 12, pp. 2094–2097, 2018

    2018

  40. [48]

    Demystifying Artificial Intelligence for Data Preparation,

    C. Chai, N. Tang, J. Fan, and Y. Luo, “Demystifying Artificial Intelligence for Data Preparation,” in International Conference on Management of Data (SIGMOD) , S. Das, I. Pandis, K. S. Candan, and S. Amer-Yahia, Eds., 2023, pp. 13–20

    2023

  41. [49]

    Machine Learning to Data Management: A Round Trip,

    L. Berti- ´Equille, A. Bonifati, and T. Milo, “Machine Learning to Data Management: A Round Trip,” inIEEE International Conference on Data Engineering (ICDE), 2018, pp. 1735–1738

    2018

  42. [50]

    Creating Embeddings of Heterogeneous Relational Datasets for Data Inte- gration Tasks,

    R. Cappuzzo, P . Papotti, and S. Thirumuruganathan, “Creating Embeddings of Heterogeneous Relational Datasets for Data Inte- gration Tasks,” in International Conference on Management of Data (SIGMOD), 2020, pp. 1335–1349

    2020

  43. [51]

    REMA: Graph Embeddings-based Relational Schema Matching,

    C. Koutras, M. Fragkoulis, A. Katsifodimos, and C. Lofi, “REMA: Graph Embeddings-based Relational Schema Matching,” in Work- shops of the EDBT/ICDT, 2020

    2020

  44. [52]

    DeepAlignment: Un- supervised Ontology Matching with Refined Word Vectors,

    P . Kolyvakis, A. Kalousis, and D. Kiritsis, “DeepAlignment: Un- supervised Ontology Matching with Refined Word Vectors,” in Conference of the North American Chapter of the Association for Com- putational Linguistics: Human Language Technologies (NAACL-HLT) , 2018, pp. 787–798

    2018

  45. [53]

    Deep Entity Matching with Pre-trained Language Models,

    Y. Li, J. Li, Y. Suhara, A. Doan, and W. Tan, “Deep Entity Matching with Pre-trained Language Models,” Proc. VLDB Endow. , vol. 14, no. 1, pp. 50–60, 2020

    2020

  46. [54]

    Deep Learning for Blocking in Entity Matching: A Design Space Exploration,

    S. Thirumuruganathan, H. Li, N. Tang, M. Ouzzani, Y. Govind, D. Paulsen, G. Fung, and A. Doan, “Deep Learning for Blocking in Entity Matching: A Design Space Exploration,” Proc. VLDB Endow., vol. 14, no. 11, pp. 2459–2472, 2021

    2021

  47. [55]

    Annotating Columns with Pre-trained Language Models,

    Y. Suhara, J. Li, Y. Li, D. Zhang, C ¸ . Demiralp, C. Chen, and W. Tan, “Annotating Columns with Pre-trained Language Models,” in International Conference on Management of Data (SIGMOD) , 2022, pp. 1493–1503

    2022

  48. [56]

    Cost-Effective In-Context Learning for Entity Resolution: A De- sign Space Exploration,

    M. Fan, X. Han, J. Fan, C. Chai, N. Tang, G. Li, and X. Du, “Cost-Effective In-Context Learning for Entity Resolution: A De- sign Space Exploration,” in IEEE International Conference on Data Engineering, (ICDE), 2024, pp. 3696–3709

    2024

  49. [57]

    BoostER: Leveraging Large Language Models for Enhancing Entity Resolu- tion,

    H. Li, S. Li, F. Hao, C. J. Zhang, Y. Song, and L. Chen, “BoostER: Leveraging Large Language Models for Enhancing Entity Resolu- tion,” in ACM Web Conference (WWW), 2024, pp. 1043–1046

    2024

  50. [58]

    Using ChatGPT for Entity Matching,

    R. Peeters and C. Bizer, “Using ChatGPT for Entity Matching,” in New Trends in Database and Information Systems (ADBIS) , 2023, pp. 221–230

    2023

  51. [59]

    Can Foundation Models Wrangle Your Data?

    A. Narayan, I. Chami, L. J. Orr, and C. R ´e, “Can Foundation Models Wrangle Your Data?” Proc. VLDB Endow. , vol. 16, no. 4, pp. 738–746, 2022

    2022

  52. [60]

    Directions Towards Efficient and Automated Data Wrangling with Large Language Models,

    Z. Zhang, P . Groth, I. Calixto, and S. Schelter, “Directions Towards Efficient and Automated Data Wrangling with Large Language Models,” in International Conference on Data Engineering Workshops (ICDEW), 2024, pp. 301–304

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.