pith. machine review for the scientific record. sign in

arxiv: 2605.05476 · v1 · submitted 2026-05-06 · 💻 cs.LG · cs.AI· cs.CL

Recognition: unknown

A Unified Benchmark for Evaluating Knowledge Graph Construction Methods and Graph Neural Networks

Fabrice Guillet, Hideaki Takeda, Mounira Harzallah, Othmane Kabal, Ryutaro Ichise

Pith reviewed 2026-05-08 17:05 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.CL
keywords knowledge graph constructiongraph neural networksbenchmarkbiomedical knowledge graphssemi-supervised node classificationgraph quality evaluation
0
0 comments X

The pith

A benchmark built from one biomedical text corpus supplies noisy auto-extracted graphs plus an expert reference to compare construction methods and measure GNN robustness on node classification.

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

The paper introduces a dual-purpose benchmark that holds graph construction methods and graph neural networks to the same evaluation standard. It starts from a single textual corpus in the biomedical domain and produces two automatically extracted graphs using different methods, together with a high-quality reference graph created by experts. The expert graph functions as an upper performance bound. By running the same semi-supervised node classification task on all three graphs inside a standardized, reproducible framework, the design isolates the contribution of graph quality from the contribution of the learning model. This setup directly addresses the difficulty of telling whether poor GNN results come from noisy input graphs or from the model itself.

Core claim

The benchmark is built in the biomedical domain from a single textual corpus and includes two automatically constructed graphs generated using different extraction methods, alongside a high-quality reference graph curated by experts that serves as an upper performance bound. This design enables controlled comparison of construction methods and systematic evaluation of GNN robustness through semi-supervised node classification, with a standardized, reproducible, and extensible evaluation framework that supports new extraction methods and learning models.

What carries the argument

The dual-purpose benchmark that pairs two automatically extracted knowledge graphs and one expert-curated reference graph, all derived from the identical textual corpus, and evaluates them jointly via semi-supervised node classification inside a fixed framework.

If this is right

  • Different graph construction methods can be ranked by how much they degrade downstream GNN accuracy on the shared node-classification task.
  • GNN robustness to noise, fragmentation, and semantic inconsistencies can be measured by comparing performance on the auto-extracted graphs against the expert reference.
  • New extraction methods or learning models can be plugged into the same evaluation pipeline without redesigning the test.
  • The gap between auto-extracted and expert-graph performance provides a concrete quantitative target for improving construction pipelines.

Where Pith is reading between the lines

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

  • The single-corpus design makes it possible to trace specific extraction errors directly to drops in node-classification accuracy.
  • If the benchmark is extended to additional domains, the same controlled comparison could reveal whether biomedical graphs are unusually sensitive to construction noise.
  • The framework could be used to test whether certain GNN architectures are more tolerant than others to the particular noise patterns that text-extraction methods introduce.

Load-bearing premise

The single textual corpus together with its expert-curated reference graph supplies a representative upper-bound baseline that generalizes to other real-world knowledge-graph use cases.

What would settle it

Running the same semi-supervised node classification task on the three graphs and finding that the expert reference graph does not produce the highest accuracy, or that the relative ordering of the two auto-extracted graphs changes when the underlying corpus is replaced.

Figures

Figures reproduced from arXiv: 2605.05476 by Fabrice Guillet, Hideaki Takeda, Mounira Harzallah, Othmane Kabal, Ryutaro Ichise.

Figure 1
Figure 1. Figure 1: Overview of the Benchmark Construction Methodology. with a canonical name and lexical variants, and is typed according to the Semantic Network, as illustrated in view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the relationship between the Semantic Network and the Metathesaurus, adopted from Bodenreider (2004), together with examples of canonical names and lexical variations for biomedical concepts. In the Semantic Network, solid arrows represent is-a relations, while dashed arrows denote other semantic relations between semantic types. At the Metathesaurus level, edges correspond to relations bet… view at source ↗
Figure 3
Figure 3. Figure 3: Stepwise construction of the reference graph (UMLS-NCI), showing the graph statistics after each filtering and preprocessing operation enabling meaningful comparison and controlled evaluation across graph construction methods. The MedMentions (MM) dataset Mohan and Li (2019), designed specifically for biomedical concept recognition, is well-suited for this purpose. It consists of 4,392 PubMed † abstracts, … view at source ↗
Figure 4
Figure 4. Figure 4: Example of a subgraph extracted from the reference graph, illustrating the local neighborhood of the node “Heart” and the semantic relations connecting it to related biomedical concepts. mention: "regulator" CUI: C0017362 Semantic type: Anatomical Structure mention: "enzymes" CUI: C1333402 Semantic type: Anatomical Structure mention: "clinical question" CUI: C2984039 Semantic type: Intellectual Product UML… view at source ↗
Figure 5
Figure 5. Figure 5: Examples of inconsistencies between MedMentions mentions and UMLS-NCI concepts, including differences in surface forms and semantic type granularity, motivating the normalization of the corpus before graph construction. LLM-based validation to filter out incorrect triples. As expected for automatically extracted knowledge graphs, the resulting structure exhibits several imperfections. Struc￾turally, it sho… view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of a subgraph of the Semantic Network schema for the selected semantic types (shown in blue) and other semantic types (shown in green), highlighting the semantic relations between them. Your KG Construction MedMentions Method corpus Reference Graph (UMLS-NCI) Data Loader Json, CSV Your GNN Model Evaluator Semi-supervised classification Leaderboard User-constructed Text-Driven KG view at source ↗
Figure 7
Figure 7. Figure 7: Overview of the evaluation pipeline. Green components correspond to user-defined inputs view at source ↗
read the original abstract

Knowledge graphs automatically constructed from text are increasingly used in real-world applications. However, their inherent noise, fragmentation, and semantic inconsistencies significantly affect the performance of Graph Neural Networks (GNNs) on downstream tasks. Assessing their performance and robustness remains difficult, as it is often unclear whether observed results stem from the learning model or from the quality of the constructed graph itself. In this work, we introduce a dual-purpose benchmark designed to jointly evaluate (i) the performance of GNNs on noisy, text-derived graphs and (ii) the effectiveness of graph construction methods on a downstream task. The benchmark is built in the biomedical domain from a single textual corpus and includes two automatically constructed graphs generated using different extraction methods, alongside a high-quality reference graph curated by experts that serves as an upper performance bound. This design enables controlled comparison of construction methods and systematic evaluation of GNN robustness through semi-supervised node classification. We further provide a standardized, reproducible, and extensible evaluation framework, facilitating the integration of new graph extraction methods and learning models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces a dual-purpose benchmark in the biomedical domain, constructed from a single textual corpus. It includes two automatically constructed knowledge graphs using different extraction methods, an expert-curated reference graph serving as an upper performance bound, and a standardized evaluation framework. The benchmark is intended to enable controlled comparisons of graph construction methods and systematic assessment of GNN robustness via semi-supervised node classification on noisy, text-derived graphs.

Significance. If empirically validated, the benchmark could provide a useful standardized resource for isolating the impact of graph construction quality (noise and fragmentation) on GNN performance in downstream tasks, addressing a recognized challenge in applying KGs to real-world domains like biomedicine. The emphasis on reproducibility and extensibility for new methods is a constructive contribution.

major comments (2)
  1. [§3] §3 (Benchmark Design): The central claim that the design 'enables controlled comparison of construction methods' and 'systematic evaluation of GNN robustness' rests on the assumption that performance gaps in semi-supervised node classification can be attributed to graph construction fidelity rather than corpus-specific artifacts, node-set mismatches, or label alignment issues. No experiments, ablations, or analyses are reported to demonstrate this isolation or to show that the expert reference is reachable by GNNs.
  2. [§4] §4 (Evaluation Framework): The manuscript treats the single-corpus, expert-curated reference as a valid upper bound without validation experiments or sensitivity analysis. This leaves open the possibility that observed differences arise from domain biases or curation subjectivity instead of the intended construction effects, undermining the benchmark's utility for the stated purpose.
minor comments (2)
  1. [Abstract] The abstract and introduction would benefit from explicit naming of the two extraction methods and the specific biomedical corpus used, to allow immediate assessment of domain specificity.
  2. [§3] Notation for graph properties (e.g., node/edge statistics across the three graphs) should be presented in a table for clarity and to support reproducibility claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our benchmark paper. We address each major comment below, clarifying the design rationale while acknowledging where additional analyses would strengthen the claims. We plan revisions to incorporate supporting experiments and discussions.

read point-by-point responses

  1. Referee: [§3] §3 (Benchmark Design): The central claim that the design 'enables controlled comparison of construction methods' and 'systematic evaluation of GNN robustness' rests on the assumption that performance gaps in semi-supervised node classification can be attributed to graph construction fidelity rather than corpus-specific artifacts, node-set mismatches, or label alignment issues. No experiments, ablations, or analyses are reported to demonstrate this isolation or to show that the expert reference is reachable by GNNs.

    Authors: We agree that explicit isolation of construction effects is necessary to fully support the benchmark's purpose. The design controls for corpus-specific artifacts by deriving all graphs from the identical textual corpus. Node sets are aligned via shared entity mentions and linking procedures, with labels standardized from the expert curation. However, we acknowledge the absence of dedicated ablations in the current manuscript. In the revision, we will add experiments such as GNN performance on the expert reference versus synthetically perturbed versions to demonstrate reachability, and controlled comparisons varying only node-set alignment or label consistency to isolate construction fidelity effects. revision: yes

  2. Referee: [§4] §4 (Evaluation Framework): The manuscript treats the single-corpus, expert-curated reference as a valid upper bound without validation experiments or sensitivity analysis. This leaves open the possibility that observed differences arise from domain biases or curation subjectivity instead of the intended construction effects, undermining the benchmark's utility for the stated purpose.

    Authors: We recognize that positioning the expert graph as an upper bound benefits from empirical validation. The reference is constructed by domain experts on the same corpus to serve as a high-quality ceiling, but we agree that sensitivity to curation subjectivity and potential domain biases requires attention. In the revised manuscript, we will incorporate validation elements, including any available inter-annotator agreement statistics from the curation process and additional GNN runs on the reference graph to illustrate achievable performance levels, thereby addressing concerns about biases versus construction effects. revision: yes

Circularity Check

0 steps flagged

No circularity: benchmark proposal contains no derivations or fitted predictions

full rationale

The paper introduces a dual-purpose benchmark for evaluating KG construction methods and GNNs on a single biomedical corpus with two auto-extracted graphs plus an expert reference. No equations, parameter fitting, predictions, or first-principles derivations appear in the abstract or described structure. The central contribution is an evaluation framework rather than a computed result that could reduce to its inputs by construction. Self-citations, if present, are not load-bearing for any claimed derivation. This matches the default expectation of no significant circularity for a benchmark paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are introduced; the paper relies on standard concepts of knowledge graphs, GNNs, and node classification without new mathematical constructs.

pith-pipeline@v0.9.0 · 5497 in / 1037 out tokens · 38296 ms · 2026-05-08T17:05:06.028724+00:00 · methodology

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

157 extracted references · 44 canonical work pages · 7 internal anchors

  1. [1]

    Advances in neural information processing systems , volume=

    Open graph benchmark: Datasets for machine learning on graphs , author=. Advances in neural information processing systems , volume=

  2. [2]

    Machine Vision and Applications , volume=

    Graph neural networks in node classification: survey and evaluation , author=. Machine Vision and Applications , volume=. 2022 , publisher=

    2022

  3. [3]

    Arrar et al

    A comprehensive survey of link prediction methods: D. Arrar et al. , author=. The journal of supercomputing , volume=. 2024 , publisher=

    2024

  4. [4]

    Neural Computing and Applications , volume=

    A survey of inductive knowledge graph completion , author=. Neural Computing and Applications , volume=. 2024 , publisher=

    2024

  5. [5]

    Nucleic acids research , volume=

    The unified medical language system (UMLS): integrating biomedical terminology , author=. Nucleic acids research , volume=. 2004 , publisher=

    2004

  6. [6]

    MedMentions: A Large Biomedical Corpus Annotated with UMLS Concepts

    Medmentions: A large biomedical corpus annotated with umls concepts , author=. arXiv preprint arXiv:1902.09476 , year=

    work page Pith review arXiv 1902

  7. [7]

    Procedia computer science , volume=

    Core-concept-seeded LDA for ontology learning , author=. Procedia computer science , volume=. 2021 , publisher=

    2021

  8. [8]

    ACM Computing Surveys , volume=

    A survey of graph neural networks for social recommender systems , author=. ACM Computing Surveys , volume=. 2024 , publisher=

    2024

  9. [9]

    Self-Supervised Learning of Graph Neural Networks: A Unified Review , year=

    Xie, Yaochen and Xu, Zhao and Zhang, Jingtun and Wang, Zhengyang and Ji, Shuiwang , journal=. Self-Supervised Learning of Graph Neural Networks: A Unified Review , year=

  10. [10]

    Computational and structural biotechnology journal , volume=

    Constructing knowledge graphs and their biomedical applications , author=. Computational and structural biotechnology journal , volume=. 2020 , publisher=

    2020

  11. [11]

    arXiv preprint arXiv:2403.04468 , year=

    A survey of graph neural networks in real world: Imbalance, noise, privacy and ood challenges , author=. arXiv preprint arXiv:2403.04468 , year=

    work page arXiv

  12. [12]

    arXiv preprint arXiv:1912.10206 , year=

    How robust are graph neural networks to structural noise? , author=. arXiv preprint arXiv:1912.10206 , year=

    work page arXiv 1912

  13. [13]

    Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Graph structure learning for robust graph neural networks , author=. Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  14. [14]

    IEEE transactions on neural networks and learning systems , volume=

    A comprehensive survey on graph neural networks , author=. IEEE transactions on neural networks and learning systems , volume=. 2020 , publisher=

    2020

  15. [15]

    Proceedings of the 26th International Joint Conference on Artificial Intelligence , pages =

    Xu, Jiacheng and Qiu, Xipeng and Chen, Kan and Huang, Xuanjing , title =. Proceedings of the 26th International Joint Conference on Artificial Intelligence , pages =. 2017 , isbn =

    2017

  16. [16]

    Artificial intelligence review , volume=

    Graph neural networks for text classification: A survey , author=. Artificial intelligence review , volume=. 2024 , publisher=

    2024

  17. [17]

    International Semantic Web Conference , pages=

    LLMs4OL: Large language models for ontology learning , author=. International Semantic Web Conference , pages=. 2023 , organization=

    2023

  18. [18]

    WordNet: an electronic lexical database , volume=

    Automated discovery of WordNet relations , author=. WordNet: an electronic lexical database , volume=. 1998 , publisher=

    1998

  19. [19]

    arXiv preprint arXiv:1806.03191 , year=

    Hearst patterns revisited: Automatic hypernym detection from large text corpora , author=. arXiv preprint arXiv:1806.03191 , year=

    work page arXiv

  20. [20]

    Toward a Method to Generate Capability Ontologies from Natural Language Descriptions , year=

    Vieira da Silva, Luis Miguel and Kocher, Aljosha and Gehlhoff, Felix and Fay, Alexander , booktitle=. Toward a Method to Generate Capability Ontologies from Natural Language Descriptions , year=

  21. [21]

    Advances in neural information processing systems , volume=

    A unified approach to interpreting model predictions , author=. Advances in neural information processing systems , volume=

  22. [22]

    Proceedings of the 2022 6th International Conference on Natural Language Processing and Information Retrieval , pages =

    Schopf, Tim and Braun, Daniel and Matthes, Florian , title =. Proceedings of the 2022 6th International Conference on Natural Language Processing and Information Retrieval , pages =. 2023 , isbn =. doi:10.1145/3582768.3582795 , abstract =

    work page doi:10.1145/3582768.3582795 2022

  23. [23]

    Core-Concept-Seeded LDA for Ontology Learning , journal =

    Hao Huang and Mounira Harzallah and Fabrice Guillet and Ziwei Xu , keywords =. Core-Concept-Seeded LDA for Ontology Learning , journal =. 2021 , note =. doi:https://doi.org/10.1016/j.procs.2021.08.023 , url =

    work page doi:10.1016/j.procs.2021.08.023 2021

  24. [24]

    In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Com- putational Linguistics: Human Language Technologies (NAACL-HLT 2019), pp

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina. BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019. doi:10.18653/v...

    work page doi:10.18653/v1/n19-1423 2019

  25. [25]

    arXiv preprint arXiv:2204.00391 , year=

    Automatic biomedical term clustering by learning fine-grained term representations , author=. arXiv preprint arXiv:2204.00391 , year=

    work page arXiv

  26. [26]

    RoBERTa: A Robustly Optimized BERT Pretraining Approach

    Roberta: A robustly optimized bert pretraining approach , author=. arXiv preprint arXiv:1907.11692 , year=

    work page internal anchor Pith review arXiv 1907

  27. [27]

    Wikidata: a free collaborative knowledgebase

    Vrande. Wikidata: a free collaborative knowledgebase , year =. Commun. ACM , month = sep, pages =. doi:10.1145/2629489 , abstract =

    work page doi:10.1145/2629489

  28. [28]

    Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Arnetminer: extraction and mining of academic social networks , author=. Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  29. [29]

    AI magazine , volume=

    Collective classification in network data , author=. AI magazine , volume=

  30. [30]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Defending graph convolutional networks against dynamic graph perturbations via bayesian self-supervision , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  31. [31]

    International Conference on Database Systems for Advanced Applications , pages=

    Learning robust representation through graph adversarial contrastive learning , author=. International Conference on Database Systems for Advanced Applications , pages=. 2022 , organization=

    2022

  32. [32]

    Proceedings of the 3rd workshop on continuous vector space models and their compositionality , pages=

    Observed versus latent features for knowledge base and text inference , author=. Proceedings of the 3rd workshop on continuous vector space models and their compositionality , pages=

  33. [33]

    ACM Computing Surveys , volume=

    A comprehensive survey on automatic knowledge graph construction , author=. ACM Computing Surveys , volume=. 2023 , publisher=

    2023

  34. [34]

    and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy

    Liu, Nelson F. and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy. Lost in the Middle: How Language Models Use Long Contexts. Transactions of the Association for Computational Linguistics. 2024. doi:10.1162/tacl_a_00638

    work page doi:10.1162/tacl_a_00638 2024

  35. [35]

    Lost in the Middle: How Language Models Use Long Contexts

    Lost in the middle: How language models use long contexts , author=. arXiv preprint arXiv:2307.03172 , year=

    work page internal anchor Pith review arXiv

  36. [36]

    Vipula Rawte, Amit Sheth, and Amitava Das

    Hallucination is inevitable: An innate limitation of large language models , author=. arXiv preprint arXiv:2401.11817 , year=

    work page arXiv

  37. [37]

    arXiv preprint arXiv:2305.03132 , year=

    The role of global and local context in named entity recognition , author=. arXiv preprint arXiv:2305.03132 , year=

    work page arXiv

  38. [38]

    Efficient Estimation of Word Representations in Vector Space

    Efficient estimation of word representations in vector space , author=. arXiv preprint arXiv:1301.3781 , year=

    work page internal anchor Pith review arXiv

  39. [39]

    Artificial intelligence , volume=

    YAGO2: A spatially and temporally enhanced knowledge base from Wikipedia , author=. Artificial intelligence , volume=. 2013 , publisher=

    2013

  40. [40]

    international semantic web conference , pages=

    Dbpedia: A nucleus for a web of open data , author=. international semantic web conference , pages=. 2007 , organization=

    2007

  41. [41]

    International conference on machine learning , pages=

    A simple framework for contrastive learning of visual representations , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  42. [42]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Bert: Pre-training of deep bidirectional transformers for language understanding , author=. arXiv preprint arXiv:1810.04805 , year=

    work page internal anchor Pith review arXiv

  43. [43]

    Kipf et al.Variational Graph Auto- Encoders

    Variational graph auto-encoders , author=. arXiv preprint arXiv:1611.07308 , year=

    work page arXiv

  44. [44]

    2014 , eprint=

    Generative Adversarial Networks , author=. 2014 , eprint=

    2014

  45. [45]

    IEEE transactions on cybernetics , volume=

    Learning graph embedding with adversarial training methods , author=. IEEE transactions on cybernetics , volume=. 2019 , publisher=

    2019

  46. [46]

    Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages=

    Mgae: Marginalized graph autoencoder for graph clustering , author=. Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages=

    2017

  47. [47]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Symmetric graph convolutional autoencoder for unsupervised graph representation learning , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  48. [48]

    Knowledge-Based Systems , volume=

    Dual-decoder graph autoencoder for unsupervised graph representation learning , author=. Knowledge-Based Systems , volume=. 2021 , publisher=

    2021

  49. [49]

    Information Sciences , volume=

    Multi-view representation model based on graph autoencoder , author=. Information Sciences , volume=. 2023 , publisher=

    2023

  50. [50]

    Advances in neural information processing systems , volume=

    Inductive representation learning on large graphs , author=. Advances in neural information processing systems , volume=

  51. [51]

    Proceedings of the sixteenth ACM international conference on web search and data mining , pages=

    S2gae: Self-supervised graph autoencoders are generalizable learners with graph masking , author=. Proceedings of the sixteenth ACM international conference on web search and data mining , pages=

  52. [52]

    2022 , eprint=

    GraphMAE: Self-Supervised Masked Graph Autoencoders , author=. 2022 , eprint=

    2022

  53. [53]

    Knowledge graph representation learning with relation-guided aggregation and interaction , journal =

    Bin Shang and Yinliang Zhao and Jun Liu , keywords =. Knowledge graph representation learning with relation-guided aggregation and interaction , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.ipm.2024.103752 , url =

    work page doi:10.1016/j.ipm.2024.103752 2024

  54. [54]

    Multi-Relational Semantic Awareness for Knowledge Graph Completion , year=

    Song, Jiawei and Duan, Zongtao and Cao, Jianrong and Lin, Yun , booktitle=. Multi-Relational Semantic Awareness for Knowledge Graph Completion , year=

  55. [55]

    arXiv preprint arXiv:2404.08313 , year=

    The integration of semantic and structural knowledge in knowledge graph entity typing , author=. arXiv preprint arXiv:2404.08313 , year=

    work page arXiv

  56. [56]

    Martinez-Rodriguez and Ivan Lopez-Arevalo and Ana B

    Jose L. Martinez-Rodriguez and Ivan Lopez-Arevalo and Ana B. Rios-Alvarado , keywords =. OpenIE-based approach for Knowledge Graph construction from text , journal =. 2018 , issn =. doi:https://doi.org/10.1016/j.eswa.2018.07.017 , url =

    work page doi:10.1016/j.eswa.2018.07.017 2018

  57. [57]

    A Survey on Open Information Extraction from Rule-based Model to Large Language Model

    Pai, Liu and Gao, Wenyang and Dong, Wenjie and Ai, Lin and Gong, Ziwei and Huang, Songfang and Zongsheng, Li and Hoque, Ehsan and Hirschberg, Julia and Zhang, Yue. A Survey on Open Information Extraction from Rule-based Model to Large Language Model. Findings of the Association for Computational Linguistics: EMNLP 2024. 2024. doi:10.18653/v1/2024.findings...

    work page doi:10.18653/v1/2024.findings-emnlp.560 2024

  58. [58]

    O pen IE 6: I terative G rid L abeling and C oordination A nalysis for O pen I nformation E xtraction

    Kolluru, Keshav and Adlakha, Vaibhav and Aggarwal, Samarth and Mausam and Chakrabarti, Soumen. O pen IE 6: I terative G rid L abeling and C oordination A nalysis for O pen I nformation E xtraction. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020. doi:10.18653/v1/2020.emnlp-main.306

    work page doi:10.18653/v1/2020.emnlp-main.306 2020

  59. [59]

    , journal=

    Wang, Xiao and Bo, Deyu and Shi, Chuan and Fan, Shaohua and Ye, Yanfang and Yu, Philip S. , journal=. A Survey on Heterogeneous Graph Embedding: Methods, Techniques, Applications and Sources , year=

  60. [60]

    Graph Attention Networks

    Graph attention networks , author=. arXiv preprint arXiv:1710.10903 , year=

    work page internal anchor Pith review arXiv

  61. [61]

    Applied Ontology , volume=

    How to classify domain entities into top-level ontology concepts using large language models , author=. Applied Ontology , volume=. 2024 , publisher=

    2024

  62. [62]

    Kggen: Extracting knowledge graphs from plain text with language models.arXiv preprint arXiv:2502.09956, 2025

    Kggen: Extracting knowledge graphs from plain text with language models , author=. arXiv preprint arXiv:2502.09956 , year=

    work page arXiv

  63. [63]

    and Bloem, Peter and van den Berg, Rianne and Titov, Ivan and Welling, Max

    Schlichtkrull, Michael and Kipf, Thomas N. and Bloem, Peter and van den Berg, Rianne and Titov, Ivan and Welling, Max. Modeling Relational Data with Graph Convolutional Networks. The Semantic Web. 2018

    2018

  64. [64]

    Open Conference Proceedings , volume=

    Daselab at llms4ol 2024 task a: Towards term typing in ontology learning , author=. Open Conference Proceedings , volume=

    2024

  65. [65]

    Advances in neural information processing systems , volume=

    Translating embeddings for modeling multi-relational data , author=. Advances in neural information processing systems , volume=

  66. [66]

    2018 , eprint=

    Convolutional 2D Knowledge Graph Embeddings , author=. 2018 , eprint=

    2018

  67. [67]

    2015 , eprint=

    Embedding Entities and Relations for Learning and Inference in Knowledge Bases , author=. 2015 , eprint=

    2015

  68. [68]

    Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages =

    Wu, Cheng and Wang, Chaokun and Xu, Jingcao and Liu, Ziyang and Zheng, Kai and Wang, Xiaowei and Song, Yang and Gai, Kun , title =. Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages =. 2023 , isbn =. doi:10.1145/3580305.3599370 , abstract =

    work page doi:10.1145/3580305.3599370 2023

  69. [69]

    2019 ,URL =

    Deep Graph Infomax ,author =. 2019 ,URL =

    2019

  70. [70]

    A contrastive variational graph auto-encoder for node clustering , journal =

    Nairouz Mrabah and Mohamed Bouguessa and Riadh Ksantini , keywords =. A contrastive variational graph auto-encoder for node clustering , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.patcog.2023.110209 , url =

    work page doi:10.1016/j.patcog.2023.110209 2024

  71. [71]

    2020 , eprint=

    Heterogeneous Deep Graph Infomax , author=. 2020 , eprint=

    2020

  72. [72]

    DRGI: Deep Relational Graph Infomax for Knowledge Graph Completion , year=

    Liang, Shuang and Shao, Jie and Zhang, Dongyang and Zhang, Jiasheng and Cui, Bin , journal=. DRGI: Deep Relational Graph Infomax for Knowledge Graph Completion , year=

  73. [73]

    2019 , eprint=

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , author=. 2019 , eprint=

    2019

  74. [74]

    In2024 IEEE 40th International Conference on Data Engineering (ICDE)

    Y. Wang and X. Yan and C. Hu and Q. Xu and C. Yang and F. Fu and W. Zhang and H. Wang and B. Du and J. Jiang , booktitle =. Generative and Contrastive Paradigms Are Complementary for Graph Self-Supervised Learning , year =. doi:10.1109/ICDE60146.2024.00234 , url =

    work page doi:10.1109/icde60146.2024.00234 2024

  75. [75]

    ACM Trans

    Zheng, Yimei and Jia, Caiyan , title =. ACM Trans. Knowl. Discov. Data , month = apr, articleno =. 2024 , issue_date =. doi:10.1145/3649143 , abstract =

    work page doi:10.1145/3649143 2024

  76. [76]

    Description-based zero-shot fine-grained entity typing , author=. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages=

    2019

  77. [77]

    arXiv preprint arXiv:2004.01267 , year=

    MZET: Memory augmented zero-shot fine-grained named entity typing , author=. arXiv preprint arXiv:2004.01267 , year=

    work page arXiv 2004

  78. [78]

    BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension

    BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension , author=. arXiv preprint arXiv:1910.13461 , year=

    work page internal anchor Pith review arXiv 1910

  79. [79]

    arXiv e-prints , pages=

    The llama 3 herd of models , author=. arXiv e-prints , pages=

  80. [80]

    Advances in neural information processing systems , volume=

    Language models are few-shot learners , author=. Advances in neural information processing systems , volume=

Showing first 80 references.