Pith. sign in

REVIEW 4 major objections 7 minor 43 references

Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Document Clustering

T0 review · 4 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that replacing co-occurrence or KNN edges with named-entity similarity edges, while keeping LLM embeddings as node features and jointly optimizing embeddings and clusters, yields better document clusters than the…

desk verdict A genuinely new NER-based graph construction is buried under an evaluation whose cluster count comes from the ground-truth labels, so the headline unsupervised gains are not yet established. read the letter →

arxiv 2412.14867 v1 pith:KPS6CQUU submitted 2024-12-19 cs.CL

Imed Keraghel , Mohamed Nadif This is my paper · ORCID
classification cs.CL
keywords LargeLanguageModelsNamedEntityRecognitionGraphConvolutionalNetworksNodeEmbeddingClusteringDocumentLLMEmbeddings
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

This paper tries to establish that document clustering improves when the graph that connects documents is built from shared named entities instead of word co-occurrence or k-nearest-neighbour lexical similarity. The authors pair that entity-similarity adjacency matrix with LLM embeddings as node features, then run a graph-convolutional clustering objective that learns embeddings and cluster assignments together. On four corpora (BBC News, MLSUM, PubMed, Arxiv-10), the resulting pipeline, GCC*, reports higher accuracy, normalized mutual information, and adjusted Rand index than co-occurrence-based, KNN-graph, and LLM-embedding-only baselines. The reason to care is that clustering entity-heavy documents such as news requires relational links between documents that surface word statistics miss.

What carries the argument

The load-bearing object is the named-entity adjacency matrix $A_{\text{ner}}$. Its entries average the cosine similarities of same-type entity pairs between two documents, with an edge allowed only when at least $\tau = 0.9$ similarity is exceeded and at least three shared entities connect the documents. This matrix is fed, together with LLM embeddings $X_{\ell\ell m}$ from a 1536-dimensional text-embedding model, into the GCC objective: $\min \|Y_p - Y_p W W^\top\|^2 + \|Y_p W - G F\|^2$, where $Y_p = T^p X_{\ell\ell m}$ and $T$ is the normalized simple-graph-convolution matrix with self-loops; the orthogonality-constrained problem reduces to $\min \|Y_p - G F W^\top\|^2$ and is solved by alternating updates of $G$, $F$, and $W$. The machinery's work is to let entity-level semantic relations shape the graph convolution, so the learned embeddings and cluster assignments are pulled toward the structure of named entities rather than surface word overlap.

What would settle it

Run the same pipeline with k selected by a fully unsupervised internal criterion, such as the largest silhouette score or the minimum Davies-Bouldin index over the same grid, or by a dendrogram cut made before the annotator sees the true labels; if the GCC* row no longer beats GCC with the KNN graph and LLM embeddings, and no longer beats LLM-only k-means, the central claim fails. A cheaper check is to re-run Arxiv-10 with k=3 or k=4, the other partitions the dendrogram suggests, and see whether the advantage over baselines survives.

Watch

Extended reading notes

Core claim

The central claim is that a document graph whose edges encode named-entity similarity, combined with LLM-derived document embeddings and a joint embedding-and-clustering objective, yields clusters that match ground-truth classes better than any of the components alone. Concretely, an edge between two documents exists only when they share at least three same-type named entities whose Word2Vec cosine similarity exceeds a threshold of 0.9, and the edge weight averages those similarities; this named-entity adjacency matrix replaces the usual KNN or co-occurrence adjacency. Feeding this matrix together with LLM embeddings into the GCC objective, minimizing reconstruction error plus a k-means-style clustering penalty on the embedded nodes, produces the reported gains, for example an accuracy of 97.61% with a normalized mutual information of 95.12% on BBC News and an NMI of 72.42% on MLSUM, exceeding the KNN-graph variant and every other baseline in each table row. The authors also observe that using the entity graph alone with spectral clustering fails because the graph is too sparse, and that the joint graph-convolutional step is what makes the entity graph usable.

Load-bearing premise

The evaluation treats the number of clusters as chosen without using the true labels, but the choice is made by inspecting dendrograms against the known classes and is set to the true class count for Arxiv-10; label leakage here would inflate every reported gain.

Editorial extensions

If this is right

  • If GCC* is correct, then on entity-rich corpora a clustering pipeline should treat named entities as first-class relational evidence: on BBC News, accuracy rises from 95.82% with the KNN graph to 97.61% with the named-entity graph while using the same LLM embeddings.
  • The joint embedding-and-clustering step is necessary to unlock the entity graph: spectral clustering on the NER adjacency alone yields near-zero NMI on PubMed, whereas GCC* keeps the graph useful by learning cluster-friendly embeddings.
  • LLM embeddings provide the dominant baseline jump over bag-of-words representations, and the named-entity graph adds a smaller but consistent gain on top of them across all four datasets.
  • The construction is model-agnostic: with Word2Vec for entity vectors and different NER models for English and French, stronger future NER or embedding models could be substituted without changing the method.

Reading between the lines

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

  • A testable extension the paper leaves open: restricting edges to a single entity type, such as organizations, locations, or persons, may sharpen clusters for topic-specific corpora, since the current edge weight averages similarities across all same-type entities.
  • Because the number of clusters is chosen by inspecting dendrograms with the true class structure in view, and set to the true class count for Arxiv-10, the headline numbers likely represent an upper bound; a fair unsupervised comparison would fix k by an internal criterion, and only if the gains persist there would GCC* be a practical recipe for entity-rich clustering.
  • The linear-algebra equivalence used to justify the objective suggests the named-entity adjacency matrix could be plugged into other spectral or matrix-factorization clustering methods, but the paper only tests it inside GCC.
  • On PubMed the gain over LLM embeddings alone is small (65.10% versus 63.25% accuracy), so the method's advantage may concentrate in documents with many named entities, which is exactly the regime the abstract claims.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. This paper proposes GCC*, a document-clustering pipeline that modifies the GCC joint embedding/clustering method of [11] in two ways: it replaces bag-of-words features with LLM embeddings (OpenAI text-embedding-3-small) as the node feature matrix Xℓℓm, and it replaces the k-NN adjacency with a named-entity graph Aner, in which two documents are connected when they share enough entities whose Word2Vec embeddings have cosine similarity above a threshold, with edge weights given by Eq. (4). The clustering objective (Eqs. (7)-(9)), the λ=1 setting, and the alternating optimization are taken from [11]. The authors report ACC, NMI, and ARI on four corpora (BBC News, a 16,321-document five-category subset of MLSUM, PubMed, and a 10,000-document sample of Arxiv-10), claiming that GCC* on (Aner, Xℓℓm) exceeds all evaluated baselines, including KNN-graph GCC, co-occurrence-based deep clustering, and k-means on LLM embeddings, and they illustrate the learned embeddings with UMAP.

Significance. If the empirical claim were established, the contribution would be a simple, low-cost recipe for injecting entity-level relational structure into LLM-based document clustering, with a transparent pipeline (NER extraction, Word2Vec entity similarities, thresholded document graph, GCC optimization) that is easy to replicate from Algorithm 1, and a benchmark sweep that includes a French corpus and a scientific-abstract corpus, which is a strength. The paper is also candid about its protocol choices, which is what makes the evaluation problems identifiable. However, the optimization machinery is entirely inherited from [11], so the novelty is the adjacency construction and the feature choice; the central claim is therefore an empirical one, and it currently rests on an evaluation protocol that selects k with knowledge of the true labels and reports no variance or significance measures. The contribution is promising but not yet demonstrated.

major comments (4)
  1. [Section 4.4] The choice of the number of clusters k leaks the ground truth into the evaluation. The text states for Arxiv-10 that 'several partitions can be considered, with 3, 4, or 10 clusters. To remain consistent with the benchmarks and our evaluation study of GCC∗, we opted for 10 clusters,' and for the other datasets that 'the true number of clusters was easily identified,' meaning the dendrogram inspection is calibrated against the benchmark labels. Because k is a free parameter of the objective in Eqs. (7)-(9) and directly determines the partition, and because ACC, NMI, and ARI in Table 2 are computed against those same labels, the reported gains (e.g., BBC News NMI 95.12%) conflate method quality with oracle knowledge of k. This contradicts the paper's own framing ('Since our framework is unsupervised') and the abstract's unsupervised-superiority claim. The same paragraph selects p only 'once the number of classes has been set,' so the p=2/3 choice inherits the leakage. A clean protocol would either fix k to the class count for all methods and separately evaluate k-detection without labels, or report Table 2 across a range of k values; either way, a label-free k-selection experiment (e.g., against an internal criterion or on a held-out split) is required before the headline claim can be accepted.
  2. [Table 2] The reported means over 20 runs come without standard deviations or significance tests, which matters because several headline comparisons are small. On PubMed, GCC∗ (ACC 65.10, NMI 29.04, ARI 25.00) is within 0.10 pp of GCC(knn, Xℓℓm) on ACC (65.00) and exactly tied on ARI (25.00); on Arxiv-10 the ACC gap is 2.87 pp. Without a paired test (e.g., Wilcoxon over the 20 runs) or variance reporting, the claim that GCC∗ 'exceeds all other methods' is not quantitatively supported on these datasets. Please report mean ± std and significance levels for at least the main comparison rows.
  3. [Sections 3.3.3-3.3.4, Eq. (4), Algorithm 1] The threshold τ is overloaded to mean two different things. In Section 3.3.3 and Section 4.3 ('we only kept links where τ was greater than 0.9'), τ is a cosine-similarity cutoff between entity embeddings, whereas in Section 3.3.4 and Eq. (4) ('the number of shared entities Edi ∩ Edj is greater than or equal to a predefined threshold τ') τ is the minimum count of shared entities, and Algorithm 1 uses τ for the similarity comparison while hardcoding 'at least 3 common named entities.' Since Eq. (4) is the definition of the adjacency matrix Aner on which the entire method rests, the construction is ambiguous: a reader cannot determine from the paper whether the count threshold is τ (as in Eq. (4)) or 3 (as in Algorithm 1 and Section 4.3). Please introduce distinct names (e.g., τ_sim for the cosine cutoff and m for the minimum number of shared entities), and reconcile Eq. (4), Algorithm 1, and the experimental setting accordingly.
  4. [Table 1 and Section 4.3] The evaluation corpus is not the one described in Table 1. Table 1 reports MLSUM as 407,835 documents with 612 clusters, but Section 4.3 says that 'we randomly sample 10,000 Arxiv documents and 16,321 MLSUM documents from five categories' (Sport, Health, Politics, Economy, Climate); the MLSUM results in Table 2 therefore appear to be for a 16,321-document, 5-class subset, not for the corpus in Table 1, and the effective k for MLSUM is left ambiguous (5 vs 612). Similarly, the Arxiv-10 column is presumably the 10,000-document sample, yet Table 1 gives 100,000 documents. The paper should state the actual number of documents and classes used for evaluation, present the subset statistics, and clarify which k was used for each column of Table 2.
minor comments (7)
  1. [Introduction] The text cites 'Figure 4' for the KNN-vs-NER graph comparison, but that figure is Figure 1; Figure 4 is the UMAP figure in Section 4.5. The cross-reference should be corrected.
  2. [Eq. (4)] The text says the edge weight is 'proportional to the similarity of the named entities' without saying whether aij is the sum or the average of the pairwise entity similarities; also, since entities are matched across documents by cosine similarity rather than string equality, the intersection Edi ∩ Edj in Eq. (4) and the count constraint should be defined over matched entity pairs, not string-identical tokens.
  3. [Section 3.4, Eqs. (8)-(9)] The equivalence of (8) and (9) is asserted by reference to [2, 3, 17, 11] without a derivation; please provide the derivation for this objective form with Agg = T^p Xℓℓm, or cite the precise result in [11] that covers it.
  4. [Abstract and Section 4.5] The abstract's 'notably for documents rich in named entities' is not tested, as no experiment varies entity density across documents; please qualify the claim or add an ablation that varies entity density.
  5. [Table 2] Decimal formatting is inconsistent (e.g., 55.8 vs 55.80, 0.9 vs 0.90, 0.0), and the table should state explicitly that all entries are percentages.
  6. [Section 2 vs Section 4] The comparison set omits the graph-clustering and NER-clustering methods reviewed in Section 2 (e.g., DGI, MVGRL, GraphCL, BERT-Kmeans, and [6, 8]); 'exceeds all other methods' should be scoped to the evaluated baselines.
  7. [Section 4.3 and Figure 4 caption] The caption says the UMAP embeddings were 'obtained with GPT,' but the paper elsewhere states that embeddings were generated with OpenAI's text-embedding-3-small model; please use the embedding model name consistently.

Circularity Check

1 steps flagged · score 5.0 of 10

Arxiv-10 cluster count is chosen from benchmark labels in Section 4.4, making the unsupervised k-detection claim and Table 2 evaluation partially circular.

  1. fitted input called prediction [Section 4.4, 'Assessing the number of clusters k and the power p']
    "For the arXiv dataset, several partitions can be considered, with 3, 4, or 10 clusters. To remain consistent with the benchmarks and our evaluation study of GCC∗, we opted for 10 clusters."

    The paper's stated aim is to assess unsupervised detection of the true number of clusters, but for Arxiv-10 it sets k to the benchmark's class count ('to remain consistent with the benchmarks'). Since k is a free parameter in objective (7)-(9) and directly fixes the number of clusters, the ACC/NMI/ARI reported in Table 2 are obtained with label-informed k, not by a purely unsupervised procedure. The 'detected' k=10 is therefore not a prediction; it is an input supplied from the evaluation labels. This does not force the document-to-cluster assignment itself, so the circularity is partial rather than total.

full rationale

No formal derivation in the paper reduces to its inputs: the optimization objective and the (8) to (9) equivalence are imported from prior work [11] (same group), but those are reusable published results and the paper adds new graph construction and embeddings, evaluated on external datasets. The only significant circular step is k-selection: Section 4.4 admits using the benchmark class count to choose k for Arxiv-10 (and similar dendrogram inspection for the other sets), which undercuts the paper's 'unsupervised' framing. Because k is a model input rather than a fitted continuous parameter and the learned assignments are still produced by the GCC algorithm, this is a partial evaluation circularity, not a derivation that equals its input by construction. Consequently, the central claim that GCC* outperforms baselines is weakened but not logically forced.

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

The method introduces two hand-set graph-building constants (0.9 and 3), two label-informed hyperparameters (k, p), and carries forward the GCC objective from prior self-authored work. No new theoretical entities are introduced, and no falsifiable prediction outside the benchmark results is offered.

free parameters (6)
  • entity similarity threshold tau = 0.9
    Cosine similarity above 0.9 between pairs of named entities of the same type is required to keep a link; chosen by hand in Section 4.3, no ablation.
  • minimum shared entity links per document edge = 3
    Algorithm 1 and Section 4.3 require at least three similar entity pairs for an edge; no sensitivity analysis.
  • number of clusters k = varies by dataset (250/500 oversegmentation, then Ward)
    Selected in Section 4.4 using dendrogram inspection and, for Arxiv-10, explicitly chosen as 10 to match benchmarks; this uses label knowledge.
  • propagation power p = 2 or 3
    Selected by minimizing square root of cluster loss over p in 1..50 in Section 4.4, with cluster loss computed under the chosen k.
  • lambda = 1
    Taken from GCC [11]; the paper notes it could be investigated in the future.
  • PC Kmeans must-link count = 500
    Top 500 strongest links used as must-link constraints; the authors state that increasing this number reduced performance, so it was tuned on test data.
assumptions (5)
  • domain assumption GCC optimization equivalence
    The claim that (8) can be solved by alternating updates of G, F and W, and that (8) is equivalent to (9), is delegated to [11], [2], [3], and [17], all from the same research group; no proof is included here.
  • domain assumption NER quality
    The method assumes that DeBERTa, camembert-ner, and GPT-4o accurately extract named entities from the four corpora; no entity-level evaluation is reported.
  • domain assumption Word2Vec semantic similarity
    Step 3 in Section 3.3.3 assumes that cosine similarity between Word2Vec vectors of entities captures semantic relatedness, and that threshold 0.9 yields meaningful edges.
  • standard math Function Tp as low-pass filter
    The normalized graph convolution T = D_T^{-1}(I + tilde S) is adopted from SGC [40] and used unmodified.
  • domain assumption LLM embeddings as node features
    The paper assumes that text-embedding-3-small represents document semantics sufficiently for clustering; this is a choice, not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Document Clustering." pith.science (2026). https://pith.science/paper/KPS6CQUU

@misc{pith2026241214867,
  author       = {Pith},
  title        = {Pith review of: Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Document Clustering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KPS6CQUU}},
  note         = {Machine review of arXiv:2412.14867}
}
read the original abstract

Recent advances in machine learning, particularly Large Language Models (LLMs) such as BERT and GPT, provide rich contextual embeddings that improve text representation. However, current document clustering approaches often ignore the deeper relationships between named entities (NEs) and the potential of LLM embeddings. This paper proposes a novel approach that integrates Named Entity Recognition (NER) and LLM embeddings within a graph-based framework for document clustering. The method builds a graph with nodes representing documents and edges weighted by named entity similarity, optimized using a graph-convolutional network (GCN). This ensures a more effective grouping of semantically related documents. Experimental results indicate that our approach outperforms conventional co-occurrence-based methods in clustering, notably for documents rich in named entities.

Figures

Figures reproduced from arXiv: 2412.14867 by the authors.

Figure 1
Figure 1. Comparison of graph structures for the BBC News dataset. (Left) KNN-based graph with lexical similarity; [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed model pipeline: LLM-based feature extraction, NER-based graph construction, and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Comparison of dendrograms for different datasets. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: UMAP projection of the cluster embeddings obtained with GPT ( [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 35 canonical work pages

  1. [11]

    Efficient graph convolution for joint node representation learning and clustering

    FETTAL , C., L ABIOD , L., AND NADIF, M. Efficient graph convolution for joint node representation learning and clustering. In WSDM (2022), pp. 289–297

    work page 2022

  2. [1]

    Regularized bi-directional co-clustering

    AFFELDT , S., L ABIOD , L., AND NADIF , M. Regularized bi-directional co-clustering. Statistics and Computing 31, 3 (2021), 32

    work page 2021

  3. [2]

    A semi-nmf-pca unified framework for data clustering

    ALLAB , K., L ABIOD , L., AND NADIF , M. A semi-nmf-pca unified framework for data clustering. IEEE TKDE 29, 1 (2016), 2–16

    work page 2016

  4. [3]

    Simultaneous spectral data embedding and clustering

    ALLAB , K., L ABIOD , L., AND NADIF , M. Simultaneous spectral data embedding and clustering. IEEE TNNLS 29, 12 (2018), 6396–6401

    work page 2018

  5. [4]

    k-means++: The advantages of careful seeding

    ARTHUR , D., AND VASSILVITSKII , S. k-means++: The advantages of careful seeding. Tech. rep., Stanford, 2006

    work page 2006

  6. [5]

    A dendrite method for cluster analysis

    CALI ´NSKI , T., AND HARABASZ , J. A dendrite method for cluster analysis. Communications in Statistics-theory and Methods 3, 1 (1974), 1–27. 9 Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Document Clustering

    work page 1974

  7. [6]

    H., T ANG , T

    CAO, T. H., T ANG , T. M., AND CHAU, C. K. Text clustering with named entities: a model, experimentation and realization. In Data Mining: Foundations and Intelligent Paradigms. 2012, pp. 267–287

    work page 2012

  8. [7]

    L., AND BOULDIN , D

    DAVIES , D. L., AND BOULDIN , D. W. A cluster separation measure. IEEE transactions on pattern analysis and machine intelligence, 2 (1979), 224–227

    work page 1979

Show all 43 references
  1. [8]

    Analysis of named entity recognition and linking for tweets

    DERCZYNSKI , L., M AYNARD , D., R IZZO , G., V AN ERP, M., G ORRELL , G., T RONCY, R., P ETRAK , J., AND BONTCHEVA , K. Analysis of named entity recognition and linking for tweets. Information Processing & Management 51, 2 (2015), 32–49

    2015

  2. [9]

    M., T HONET , T., AND GAUSSIER , E

    FARD , M. M., T HONET , T., AND GAUSSIER , E. Deep k-means: Jointly clustering with k-means and learning representations. Pattern Recognition Letters 138 (2020), 185–192

    2020

  3. [10]

    Protoformer: Embedding prototypes for transformers

    FARHANGI , A., S UI, N., H UA, N., B AI, H., H UANG , A., AND GUO, Z. Protoformer: Embedding prototypes for transformers. In PAKDD (2022), pp. 447–458

    2022

  4. [12]

    HASSANI , K., AND KHASAHMADI , A. H. Contrastive multi-view representation learning on graphs. In ICML (2020), pp. 4116–4126

    2020

  5. [13]

    Beyond words: a comparative analysis of llm embeddings for effective clustering

    KERAGHEL , I., M ORBIEU , S., AND NADIF , M. Beyond words: a comparative analysis of llm embeddings for effective clustering. In IDA (2024), pp. 205–216

    2024

  6. [14]

    K.-S., AND NASRIDINOV , A

    KIM, J.-H., C HOI , J.-H., P ARK , Y.-H., L EUNG , C. K.-S., AND NASRIDINOV , A. Knn-sc: novel spectral clustering algorithm using k-nearest neighbors. IEEE Access 9 (2021), 152616–152627

    2021

  7. [15]

    N., AND WELLING , M

    KIPF, T. N., AND WELLING , M. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907 (2016)

    2016 arXiv

  8. [16]

    N., AND WELLING , M

    K IPF, T. N., AND WELLING , M. Variational graph auto-encoders. arXiv preprint arXiv:1611.07308 (2016)

    2016 arXiv

  9. [17]

    Efficient regularized spectral data embedding

    LABIOD , L., AND NADIF , M. Efficient regularized spectral data embedding. Advances in Data Analysis and Classification 15, 1 (2021), 99–119

    2021

  10. [18]

    Pre-training of deep bidirectional transformers for language understanding

    LEE, J., AND TOUTANOVA , K. Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805 3, 8 (2018)

    2018 arXiv

  11. [19]

    Some methods for classification and analysis of multivariate observations

    MACQUEEN , J. Some methods for classification and analysis of multivariate observations. In Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability/University of California Press(1967)

    1967

  12. [20]

    Efficient estimation of word representations in vector space

    MIKOLOV, T. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781 (2013)

    2013 arXiv

  13. [21]

    Mteb: Massive text embedding benchmark

    MUENNIGHOFF , N., T AZI , N., M AGNE , L., AND REIMERS , N. Mteb: Massive text embedding benchmark. arXiv preprint arXiv:2210.07316 (2022)

    2022 arXiv

  14. [22]

    On spectral clustering: Analysis and an algorithm

    NG, A., J ORDAN , M., AND WEISS , Y. On spectral clustering: Analysis and an algorithm. Advances in neural information processing systems 14 (2001)

    2001

  15. [23]

    L., W ANG , C.-D., G U, Z., AND LI, Y

    QIN, Y., Y U, Z. L., W ANG , C.-D., G U, Z., AND LI, Y. A novel clustering method based on hybrid k-nearest- neighbor graph. Pattern recognition 74(2018), 1–14

    2018

  16. [24]

    Improving language understanding by generative pre-training

    R ADFORD , A. Improving language understanding by generative pre-training

  17. [25]

    Sentence-bert: Sentence embeddings using siamese bert-networks

    REIMERS , N. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084 (2019)

    2019 arXiv

  18. [26]

    ROUSSEEUW , P. J. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of computational and applied mathematics 20 (1987), 53–65

    1987

  19. [27]

    Directional co-clustering

    SALAH , A., AND NADIF, M. Directional co-clustering. Advances in Data Analysis and Classification 13 (2019), 591–620

    2019

  20. [28]

    Simple and effective graph autoencoders with one-hop linear models

    SALHA , G., H ENNEQUIN , R., AND VAZIRGIANNIS , M. Simple and effective graph autoencoders with one-hop linear models. In ECML-PKDD (2021), pp. 319–334

    2021

  21. [29]

    D., AND RAGHAVAN, P

    SCHÜTZE , H., M ANNING , C. D., AND RAGHAVAN, P. Introduction to information retrieval, vol. 39. Cambridge University Press Cambridge, 2008

    2008

  22. [30]

    Mlsum: The multilingual summarization corpus

    SCIALOM , T., D RAY, P.-A., L AMPRIER , S., P IWOWARSKI , B., AND STAIANO , J. Mlsum: The multilingual summarization corpus. arXiv preprint arXiv:2004.14900 (2020)

    2020 arXiv

  23. [31]

    SIA, S., D ALMIA , A., AND MIELKE , S. J. Tired of topic models? clusters of pretrained word embeddings make for fast and good topics too! arXiv preprint arXiv:2004.14914 (2020). 10 Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Do...

    2020 arXiv

  24. [32]

    Properties of the hubert-arable adjusted rand index

    S TEINLEY , D. Properties of the hubert-arable adjusted rand index. Psychological methods 9, 3 (2004), 386

    2004

  25. [33]

    Cluster ensembles—a knowledge reuse framework for combining multiple partitions

    STREHL , A., AND GHOSH , J. Cluster ensembles—a knowledge reuse framework for combining multiple partitions. JMLR 3 (2002), 583–617

    2002

  26. [34]

    T-NER: An all-round python library for transformer-based named entity recognition

    USHIO , A., AND CAMACHO -C OLLADOS , J. T-NER: An all-round python library for transformer-based named entity recognition. In EACL (2021), pp. 53–62

    2021

  27. [35]

    Attention is all you need

    V ASWANI , A. Attention is all you need. Advances in Neural Information Processing Systems (2017)

    2017

  28. [36]

    L., L IÒ, P., B ENGIO , Y., AND HJELM , R

    VELICKOVIC , P., F EDUS , W., H AMILTON , W. L., L IÒ, P., B ENGIO , Y., AND HJELM , R. D. Deep graph infomax. ICLR (Poster) 2, 3 (2019), 4

    2019

  29. [37]

    Structural deep network embedding

    W ANG , D., C UI, P., AND ZHU, W. Structural deep network embedding. In SIGKDD (2016), pp. 1225–1234

    2016

  30. [38]

    WARD JR, J. H. Hierarchical grouping to optimize an objective function. Journal of the American statistical association 58, 301 (1963), 236–244

    1963

  31. [39]

    A semantic approach for text clustering using wordnet and lexical chains

    WEI, T., L U, Y., C HANG , H., Z HOU , Q., AND BAO, X. A semantic approach for text clustering using wordnet and lexical chains. Expert Systems with applications 42, 4 (2015), 2264–2275

    2015

  32. [40]

    Simplifying graph convolutional networks

    WU, F., S OUZA , A., Z HANG , T., F IFTY, C., Y U, T., AND WEINBERGER , K. Simplifying graph convolutional networks. In ICML (2019), pp. 6861–6871

    2019

  33. [41]

    D., AND HONG , M

    YANG , B., F U, X., S IDIROPOULOS , N. D., AND HONG , M. Towards k-means-friendly spaces: Simultaneous deep learning and clustering. In ICML (2017), pp. 3861–3870

    2017

  34. [42]

    Graph contrastive learning with augmenta- tions

    Y OU, Y., C HEN , T., S UI, Y., C HEN , T., WANG , Z., AND SHEN , Y. Graph contrastive learning with augmenta- tions. Advances in neural information processing systems 33 (2020), 5812–5823

    2020

  35. [43]

    Graph-bert: Only attention is needed for learning graph representations

    ZHANG , J., Z HANG , H., X IA, C., AND SUN, L. Graph-bert: Only attention is needed for learning graph representations. arXiv preprint arXiv:2001.05140 (2020). 11

    2020 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.