Pith. sign in

REVIEW 6 major objections 4 minor 32 references

Un cadre paraconsistant pour l'{\'e}valuation de similarit{\'e} dans les bases de connaissances

T0 review · 6 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A paraconsistent similarity measure S* subtracts a contradiction ratio from a shared-property ratio, so conflicting knowledge pairs can score negative while staying reflexive and symmetric.

desk verdict A genuine but elementary S* formula surrounded by a broken repair pipeline; reject as is, but salvageable as a short note. read the letter →

arxiv 2509.08433 v1 pith:BMEXWDLG submitted 2025-09-10 cs.DB cs.ITcs.LOcs.SCmath.CTmath.IT

José-Luis Vilchis Medina (ENSTA Bretagne , Lab-STICC , Lab-STICC_ROBEX) This is my paper · ORCID
classification cs.DBcs.ITcs.LOcs.SCmath.CTmath.IT MSC 03B5368T2768T30
keywords similaritymeasureparaconsistentlogicknowledgebasescontradictionhandlingJaccardhierarchicalclusteringrepairmulti-agentsystems
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 argues that similarity between knowledge-base entities should account for contradictions rather than ignore them. It defines S*(K1,K2) as the fraction of shared properties minus the fraction of contradictory property pairs, so a pair that shares one trait but clashes on two gets a negative score. The measure is reflexive, symmetric, bounded between -1 and 1, and collapses to a Jaccard-style overlap ratio when no contradictions are present. A contradiction extractor and a minimal-repair mechanism are added so inconsistent entities can be compared after removing only what is needed to restore coherence. The practical stakes: in medicine, law, or multi-agent systems, two descriptions can be superficially close yet logically opposed, and a scalar measure that encodes that opposition is more honest than one that ignores it.

What carries the argument

The load-bearing object is the pair of ratios S+ and D± sharing the same denominator |Ptotal|. S+ = |Pshared|/|Ptotal| rewards common ground; D± = |Pcontradictory|/|Ptotal| charges for contradictions. Their difference S* turns similarity into a signed quantity: positive means agreement dominates, zero means a tie, and negative means conflict dominates. The shared denominator keeps the two terms comparable, and when Pcontradictory is empty S* becomes a literal-level Jaccard ratio. Supporting machinery includes the contradiction extractor E, which identifies contradictory literals, the minimal repair Rmin, which removes the smallest set needed to restore consistency, and the super-category def

What would settle it

Recompute S*(K1,K'2) for Example 2: K1 = {fièvre, toux, ¬maux de tête} and K'2 = {fièvre, maux de tête}. Shared = {fièvre}; contradictory = {maux de tête}; Ptotal = {fièvre, toux, ¬maux de tête, maux de tête}, so S* = 1/4 − 1/4 = 0, not 0.25 as printed. Also check whether E(K2) for K2 = {fièvre, ¬toux, maux de tête} alone can contain toux; under the single-entity Definition 4 it cannot, so the extraction step used in the example is not reproducible.

Watch

Extended reading notes

Core claim

The paper's discovery is that classical overlap-based similarity can be extended without abandoning its simple set-ratio form: replace 'shared over total' by the difference between the shared-property ratio and the contradictory-property ratio. With Ptotal the set of literals appearing in either entity, Pshared the literals occurring in both, and Pcontradictory the contradictory pairs, S* = |Pshared|/|Ptotal| − |Pcontradictory|/|Ptotal|. The paper proves reflexivity, symmetry, and the [-1,1] bounds, and shows on medical-style examples how negative scores separate logically opposed entities. It also defines paraconsistent super-categories as threshold-based clusters and adds a contradiction e

Load-bearing premise

The repair mechanism assumes a contradiction extractor that can see contradictions between two entities being compared, but the formal definition only lets it look inside one entity; if it cannot pair literals across entities, the repair step has no defined input.

Editorial extensions

If this is right

  • A pair that shares one property and contradicts on two gets S* = −1/5 in the paper's example, so logical conflict can outweigh overlap in the final score.
  • When no contradictory pair exists, S* equals |Pshared|/|Ptotal|, making the measure backward-compatible with Jaccard-like similarity on consistent knowledge.
  • S*(K,K) = 1 and S*(K1,K2) = S*(K2,K1), so clustering by thresholds is well-behaved and does not depend on the order in which entities are compared.
  • After applying minimal repair Rmin, the repaired entities are compared with the same formula, yielding a consistency-restored similarity score.
  • Super-categories group entities with S* > θ and separate different groups at S* ≤ θ, producing a dynamic hierarchy from a single threshold parameter.

Reading between the lines

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

  • A probabilistic or fuzzy extension of S* could replace exact counts by expected cardinalities, making the same ratio-difference interpretable when property membership is uncertain rather than definite.
  • In multi-agent systems, the sign of S* could act as a conflict signal: negative scores flag pairs that need repair or negotiation before cooperation, an application the paper names but does not develop in detail.
  • The threshold-θ hierarchy could drive active data collection: pairs scoring near θ are exactly those whose contradictions most affect the clustering, so querying them first would sharpen the hierarchy at minimal cost.
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

6 major / 4 minor

Summary. The paper proposes a paraconsistent framework for comparing knowledge-base entities represented as sets of literals. Its main measure is S*(K1,K2) = (|Pshared| - |Pcontradictory|) / |Ptotal| (Eqs. 3-4), intended to reward shared properties and penalize contradictions. The framework also defines paraconsistent super-categories Ξ*_K (Eq. 9), a contradiction extractor E (Eq. 11), and a repair mechanism, and it claims reflexivity, symmetry, boundedness, category preservation, and post-repair coherence. Two worked examples (Section 3.7) and a comparison with Jaccard similarity are provided. The paper is written in French and has an abstract, contributions, theory, examples, discussion, and conclusion.

Significance. If the basic S* formula were the only contribution, it would be a simple, interpretable, parameter-free normalized difference that reduces to Jaccard when no contradictory pairs are present. However, the additional advertised contributions—the hierarchical super-categories, the contradiction extractor, and the repair mechanism—are not correctly defined or demonstrated. The formal results are either immediate consequences of the chosen normalization or are unsupported. There is no empirical validation, and Section 4.3 explicitly concedes that validation on real datasets remains to be done. The underlying idea is salvageable, but in its current form the manuscript does not support the main claims.

major comments (6)
  1. [§3.3, Definition 2 and Proposition 1] The sets Ptotal and Pcontradictory are never formally defined, and Proposition 1 (Eq. 5) is false for inconsistent entities. If K={p,¬p}, comparing K with itself yields at least one contradictory pair (p,¬p) under Definition 1, so Pcontradictory≠∅ and S*(K,K)≠1. The proof 'par définition, Pcontradictory=∅' is not supported by any definition and conflicts with the existence of internal contradictions recognized by E in Eq. (11). Reflexivity cannot be guaranteed as stated.
  2. [§3.5, Definition 3 (Eq. 9)] The definition of Ξ*_K includes Ki only if S*(Ki,Kj)>θ for all j≠i. In Example 2, K1 and K3 do satisfy S*(K1,K3)=0.5>0.4, but K1 fails against K4 (S*=-0.17) and against K2, so under Eq. (9) neither K1 nor K3 belongs to any super-category. The displayed partition {{K1,K3},{K2},{K4},{K5}} (Eq. 17) is not the set defined by Eq. (9). Either the quantifier is wrong (e.g., intended ∃j≠i or a transitive-closure rule) or the example is inconsistent. Theorem 1 also lacks a precise definition of 'different super-categories' under Eq. (9).
  3. [§3.6.1 and Example 2 (Eq. 18)] Definition 4 defines E as an intra-entity operator: E(K) is the set of pi∈K contradicting some qj∈K. In Example 2, E(K2) is claimed to be {toux, ¬toux}, but toux∉K2 and K2 contains no complementary pair, so Eq. (11) gives E(K2)=∅. Detecting the contradiction between K1 and K2 requires a pair-level extractor, which is never defined. Hence the repair step 'supprimons ¬toux de K2' is not licensed by the framework.
  4. [§3.7, Eq. (19)] The only numerical repair calculation is arithmetically wrong. With K1={fièvre,toux,¬maux de tête} and K'_2={fièvre,maux de tête}, the shared set is {fièvre}, the contradictory pair (¬maux de tête / maux de tête) remains, and the union has 4 literals. Hence S*(K1,K'_2)=(1-1)/4=0, not 0.25. The value 0.25 ignores the remaining contradiction. This invalidates the illustrative support for the repair mechanism and, consequently, for Proposition 5, Theorem 2, and Corollary 3.1.
  5. [§3.6.5, Proposition 5] The characterization 'K réparable si et seulement si E(K)≠K' is false as stated. If E(K)=K, e.g., K={p,¬p}, choosing R=K yields K\R=∅, which is contradiction-free; Definition 5 is satisfied. The proof's claim that E(K)=K leaves no coherent subset is incorrect because the empty subset is coherent under the formal definition. The minimal-repair claims (Corollary 1.1, Corollary 3.1) therefore lack a valid basis.
  6. [§3.6.5, Theorems 2-3 and Corollary 3.2] These results are asserted rather than proved. Theorem 2 claims that repaired similarity preserves original similarity categories, but in Example 2 the category changes from negative (S*=-0.2) to S*=0 (or claimed 0.25) after repair; the categories are not preserved. Theorem 3's proof is a restatement and does not show how repairing a super-category yields coherence in 'all sub-categories.' Corollary 3.2 inherits this gap. These should be stated as conjectures or proved from the repair definition.
minor comments (4)
  1. [§3.7, Example 2] For pairs with no overlap and no contradiction, the denominator is given as 5 instead of 6: (K1,K5), (K2,K5), and (K3,K5). The scores are 0 either way, but the computed fractions are inconsistent with the definition of Ptotal.
  2. [§4.3] The paper concedes that 'une évaluation sur des ensembles de données réels restent à réaliser,' which contradicts the abstract's and conclusion's wording that the framework has been validated and is promising for practical applications.
  3. [Abstract and conclusion] The abstract and conclusion claim validation by examples from medical and legal domains, but Section 3.7 contains only medical examples; no legal example is provided.
  4. [Introduction and Definition 6] The citation key [?] in the first paragraph of the Introduction is unresolved. Additionally, Definition 6 equates ΞRP(K1,K2) with ΞP(K'_1,K'_2), mixing a similarity measure with a property-space notation; this type mismatch makes the definition hard to interpret. The notation K'≈, K'≠ in Theorem 2 is not defined.

Circularity Check

3 steps flagged · score 6.0 of 10

The advertised guarantees are largely built into the definitions: S*'s penalization behavior is Eq. (3)-(4) itself, Propositions 1-3 are immediate substitutions, and the repair-preservation theorem is asserted via Definition 6 rather than proven; the only repair example is undefined and arithmetically inconsistent.

  1. self definitional [Section 3.3, Definition 2, Eqs. (3)-(4); Section 3.4, Proposition 1]
    "S∗(K1, K2) = S+(K1, K2) − D±(K1, K2), où : S+(K1,K2)=|Pshared|/|Ptotal|, D±(K1,K2)=|Pcontradictory|/|Ptotal|. ... Cette formulation permet de pénaliser les contradictions tout en valorisant les propriétés partagées."

    The claimed behavioral achievement ('pénaliser les contradictions tout en valorisant les propriétés partagées') is not derived from an independent model; it is literally the defining formula: add the shared-property fraction and subtract the contradictory-property fraction. Propositions 1-3 are likewise immediate by substitution: for K=K, Pshared=Ptotal and Pcontradictory=∅, so S*=1; symmetry and bounds follow from the symmetric identical form of the two terms and normalization by |Ptotal|. These are properties built into the normalization, not independent theoretical results. The paper presents them as guarantees, but they reduce by construction to the definition.

  2. self definitional [Section 3.6.5, Théorème 2 and its demonstration]
    "Théorème 2 (Préservation des catégories de similarité). La similarité paraconsistante avec réparations ΞRP(K1,K2) préserve les catégories de similarité originales K′≈, K′≠ après réparation de K1 et K2. Démonstration. D’après la définition 6, ΞRP(K1,K2) est calculée sur les versions réparées K′1 et K′2. Puisque les réparations résolvent les contradictions sans altérer les propriétés partagées, les catégories de similarité restent cohérentes avec l’original ΞP(K1,K2)."

    The theorem's conclusion—that repaired similarity preserves the original similarity categories—is exactly what the proof assumes in the phrase 'sans altérer les propriétés partagées.' Definition 6 only stipulates that ΞRP(K1,K2) is the ordinary ΞP of the repaired entities; it says nothing comparing that value to the original ΞP(K1,K2). No argument links K′1,K′2 to K1,K2 beyond the unproved assertion that shared properties are unchanged. Thus the central preservation guarantee is circular: it is an input assumption dressed as a theorem.

1 more flagged steps
  1. other [Section 3.6.1, Definition 4, Eq. (11) vs. Section 3.7, Example 2, Eq. (18)]
    "E(K) = {pi ∈ K | ∃qj ∈ K tels que (pi ∧ ¬qj) ∨ (¬pi ∧ qj)}. ... E(K1) = ∅, E(K2) = {toux, ¬toux}."

    Definition 4 quantifies only within a single entity K. Applied to K2={fièvre, ¬toux, maux de tête}, there is no complementary pair inside K2, so the defined E(K2)=∅. The example nevertheless reports E(K2)={toux, ¬toux}, importing the literal toux from K1 into E(K2) without defining any pair-level extractor. The subsequent repair and recomputation S*(K1,K′2)=0.25 in Eq. (19) are therefore not consequences of the stated framework; they are an ad hoc construction. The arithmetic is also inconsistent with Eqs. (3)-(4): the retained contradiction ¬maux de tête / maux de tête gives S*(K1,K′2)=0, not 0.25. The repair pipeline is thus undefined and its only demonstration is self-inconsistent.

full rationale

The bare formula S* = |Pshared|/|Ptotal| − |Pcontradictory|/|Ptotal| is a definition, and, as such, its stated behavior and the reflexivity/symmetry/boundedness propositions follow by substitution; this is a tautology rather than a derived prediction, but it does not by itself make the paper circular in a damaging sense. The more serious issue is the contradiction-management apparatus advertised as a core contribution: Theorem 2's proof of preservation of similarity categories reduces to the assertion that repair preserves shared properties—exactly the statement to be proven—while Definition 6 provides no bridge between the repaired score and the original categories. Independently, the extractor E is formally defined only for a single entity, yet Example 2 applies it to a pair and produces E(K2)={toux, ¬toux} although K2 contains no positive toux literal; the repair step depends on this undefined cross-entity behavior, and the one reported repaired score is arithmetically inconsistent with Eq. (3)-(4). Section 4.3 also concedes that real-data validation remains to be done, undercutting the abstract/conclusion's claimed validation. There is no load-bearing self-citation chain and no fitted-data prediction, so the score is not higher; but the central consistency-preservation and repair claims are partially circular/unsupported, giving a score of 6.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The framework rests on a FOL literal representation, a silently chosen union normalization for Ptotal, an undefined coherence predicate, and an assumption that deleting contradictory literals preserves similarity categories. No free parameters are fitted to data; the only user-chosen constant is the clustering threshold theta (0.4 in the example). No new physical or logical entities with independent falsifiable handles are introduced.

free parameters (1)
  • similarity threshold theta = 0.4 in Example 2
    User-chosen threshold for forming super-categories Xi*_K; no procedure for selecting it is given, and it is not fitted to data.
assumptions (4)
  • domain assumption Knowledge entities are finite sets of closed first-order literals, so contradictions appear as complementary pairs (p, ¬p).
    Used in Section 3.1 and throughout; excludes non-logical or uncertain properties and assumes every claim has a recognizable negation.
  • ad hoc to paper Ptotal counts the union of signed literals across both entities, including p and ¬p as separate elements.
    Needed for the bound S* in [-1,1]; introduced silently in the examples rather than formally defined before Equation 4.
  • ad hoc to paper Deleting contradictory literals from an entity does not alter the shared-property counts that determine similarity categories.
    Assumed in Theorem 2, Theorem 3, and Example 2; the example's own arithmetic contradicts it when recomputed correctly.
  • domain assumption Contradictoire(K) is a well-defined predicate on literal sets.
    Used in Definition 5 and Corollary 1.1; never defined, and its decidability is not discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Un cadre paraconsistant pour l'{\'e}valuation de similarit{\'e} dans les bases de connaissances." pith.science (2026). https://pith.science/paper/BMEXWDLG

@misc{pith2026250908433,
  author       = {Pith},
  title        = {Pith review of: Un cadre paraconsistant pour l'\'evaluation de similarit\'e dans les bases de connaissances},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BMEXWDLG}},
  note         = {Machine review of arXiv:2509.08433}
}
abstract

This article proposes a paraconsistent framework for evaluating similarity in knowledge bases. Unlike classical approaches, this framework explicitly integrates contradictions, enabling a more robust and interpretable similarity measure. A new measure $ S^* $ is introduced, which penalizes inconsistencies while rewarding shared properties. Paraconsistent super-categories $ \Xi_K^* $ are defined to hierarchically organize knowledge entities. The model also includes a contradiction extractor $ E $ and a repair mechanism, ensuring consistency in the evaluations. Theoretical results guarantee reflexivity, symmetry, and boundedness of $ S^* $. This approach offers a promising solution for managing conflicting knowledge, with perspectives in multi-agent systems.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 30 canonical work pages

  1. [1]

    Probabilistic logic under co- herence, model-theoretic probabilistic logic, and de- fault reasoning in system p

    Veronica Biazzo, Angelo Gilio, Thomas Lukasiewicz, and Giuseppe Sanfilippo. Probabilistic logic under co- herence, model-theoretic probabilistic logic, and de- fault reasoning in system p. Journal of Applied Non- Classical Logics, 12(2) :189–213, 2002

    work page 2002

  2. [2]

    Théorème 3 (Cohérence après réparation)

    Puisque les réparations résolvent les contradictions sans altérer les pro- priétés partagées, les catégories de similarité restent cohé- rentes avec l’original ΞP (K1, K2). Théorème 3 (Cohérence après réparation) . Si S est une super-catégorie dans ΞK, alors la réparation de S assure la cohérence dans toutes les sous-catégories. Démonstration. D’après la ...

    work page

  3. [3]

    (19) Même après réparation, K1 et K ′ 2 ne dépassent pas le seuil de regroupement

    = 1 − 0 4 = 0.25 < θ. (19) Même après réparation, K1 et K ′ 2 ne dépassent pas le seuil de regroupement. Étape 5 : Hiérarchie finale et interprétation La structure hiérarchique finale est donc : Ξ∗ K = {{K1, K3}, {K2}, {K4}, {K5}} . (20) L’interprétation clinique est claire : — {K1, K3} représente des patients présentant des symptômes similaires liés aux ...

    work page

  4. [4]

    The tractability of subsumption in frame-based description languages

    Ronald J Brachman and Hector J Levesque. The tractability of subsumption in frame-based description languages. In AAAI, volume 84, pages 34–37, 1984

    work page 1984

  5. [5]

    A mea- sure of similarity between graph vertices : Applica- tions to synonym extraction and web searching.SIAM review, 46(4) :647–666, 2004

    Vincent D Blondel, Anahí Gajardo, Maureen Hey- mans, Pierre Senellart, and Paul Van Dooren. A mea- sure of similarity between graph vertices : Applica- tions to synonym extraction and web searching.SIAM review, 46(4) :647–666, 2004

    work page 2004

  6. [6]

    Knowledge representation and reasoning

    Ronald Brachman and Hector Levesque. Knowledge representation and reasoning. Elsevier, 2004

    2004

  7. [7]

    On the theory of inconsistent formal systems

    Newton CA Da Costa. On the theory of inconsistent formal systems. Notre dame journal of formal logic , 15(4) :497–510, 1974

    work page 1974

  8. [8]

    Universal sentence encoder

    Daniel Cer, Yinfei Yang, Sheng-yi Kong, Nan Hua, Nicole Limtiaco, Rhomni St John, Noah Constant, Mario Guajardo-Cespedes, Steve Yuan, Chris Tar, et al. Universal sentence encoder. arXiv preprint arXiv :1803.11175, 2018

    arXiv 2018

Show all 32 references
  1. [9]

    Semantic mo- deling of natural scenes based on contextual bayesian networks

    Huanhuan Cheng and Runsheng Wang. Semantic mo- deling of natural scenes based on contextual bayesian networks. Pattern Recognition, 43(12) :4042–4054, 2010

    2010

  2. [10]

    A survey of graph edit distance

    Xinbo Gao, Bing Xiao, Dacheng Tao, and Xuelong Li. A survey of graph edit distance. Pattern Analysis and applications, 13 :113–129, 2010

    2010

  3. [11]

    Bert : Pre-training of deep bidi- rectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert : Pre-training of deep bidi- rectional transformers for language understanding. In Proceedings of the 2019 conference of the North Ame- rican chapter of the association for computational linguistics : human...

    2019

  4. [12]

    Ontology matching : A machine lear- ning approach

    AnHai Doan, Jayant Madhavan, Pedro Domingos, and Alon Halevy. Ontology matching : A machine lear- ning approach. Handbook on ontologies, pages 385– 403, 2004

    2004

  5. [13]

    An over- view of ontoclean

    Nicola Guarino and Christopher A Welty. An over- view of ontoclean. Handbook on ontologies , pages 201–220, 2009

    2009

  6. [14]

    Goldstone and J.Y

    R.L. Goldstone and J.Y . Son. Similarity. Oxford Uni- versity Press, 2012

    2012

  7. [15]

    Description logic programs : Combi- ning logic programs with description logic

    Benjamin N Grosof, Ian Horrocks, Raphael V olz, and Stefan Decker. Description logic programs : Combi- ning logic programs with description logic. InProcee- dings of the 12th international conference on World Wide Web, pages 48–57, 2003

    2003

  8. [16]

    Incomplete data : what went wrong, and how to fix it

    Leonid Libkin. Incomplete data : what went wrong, and how to fix it. In Proceedings of the 33rd ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems, pages 1–13, 2014

    2014

  9. [17]

    Probabilistic latent semantic in- dexing

    Thomas Hofmann. Probabilistic latent semantic in- dexing. In Proceedings of the 22nd annual internatio- nal ACM SIGIR conference on Research and develop- ment in information retrieval, pages 50–57, 1999

    1999

  10. [18]

    Learning semantic similarity.Advances in neu- ral information processing systems, 15, 2002

    Jaz Kandola, Nello Cristianini, and John Shawe- taylor. Learning semantic similarity.Advances in neu- ral information processing systems, 15, 2002

    2002

  11. [19]

    Probabilistic reasoning in intelligent sys- tems : networks of plausible inference

    Judea Pearl. Probabilistic reasoning in intelligent sys- tems : networks of plausible inference. Elsevier, 2014

    2014

  12. [20]

    Some natural properties of strong- identification in inductive inference

    Eliana Minicozzi. Some natural properties of strong- identification in inductive inference. Theoretical Computer Science, 2(3) :345–360, 1976

    1976

  13. [21]

    Hamming distance metric learning

    Mohammad Norouzi, David J Fleet, and Russ R Sa- lakhutdinov. Hamming distance metric learning. Ad- vances in neural information processing systems , 25, 2012

    2012

  14. [22]

    A logic for concepts and similarity

    Mikhail Sheremet, Dmitry Tishkovsky, Frank Wolter, and Michael Zakharyaschev. A logic for concepts and similarity. Journal of Logic and Computation , 17(3) :415–452, 2007

    2007

  15. [23]

    Cependant, ces approches sont confrontées à des défis computationnels im- portants, notamment lorsqu’il s’agit de traiter de grandes bases de données

    offrent des représentations structurées des connais- sances, permettant une meilleure interprétabilité et la mo- délisation de relations complexes [2, 24]. Cependant, ces approches sont confrontées à des défis computationnels im- portants, notamment lorsqu’il s’agit de traiter...

  16. [24]

    Beyond the limits of thought

    Graham Priest. Beyond the limits of thought. Oxford University Press, 2002

    2002

  17. [25]

    Using information content to evaluate semantic similarity in a taxonomy

    Philip Resnik. Using information content to evaluate semantic similarity in a taxonomy. arXiv preprint cmp-lg/9511007, 1995

    1995 arXiv

  18. [26]

    Ontology mat- ching : state of the art and future challenges

    Pavel Shvaiko and Jérôme Euzenat. Ontology mat- ching : state of the art and future challenges. IEEE Transactions on knowledge and data engineering , 25(1) :158–176, 2011

    2011

  19. [27]

    Conceptual graphs

    John F Sowa. Conceptual graphs. Foundations of ar- tificial intelligence, 3 :213–237, 2008

    2008

  20. [28]

    From frequency to meaning : Vector space models of semantics

    Peter D Turney and Patrick Pantel. From frequency to meaning : Vector space models of semantics. Journal of artificial intelligence research, 37 :141–188, 2010

    2010

  21. [29]

    Features of similarity

    Amos Tversky. Features of similarity. Psychological review, 84(4) :327, 1977

    1977

  22. [30]

    Building intelligent da- tabases through similarity : Interaction of logical and qualitative reasoning

    José-Luis Vilchis-Medina. Building intelligent da- tabases through similarity : Interaction of logical and qualitative reasoning. In 2024 7th International Conference on Algorithms, Computing and Artificial Intelligence (ACAI), pages 1–5, 2024

    2024

  23. [31]

    Efficient approximate entity matching using jaro-winkler dis- tance

    Yaoshu Wang, Jianbin Qin, and Wei Wang. Efficient approximate entity matching using jaro-winkler dis- tance. In International conference on web information systems engineering, pages 231–239. Springer, 2017

    2017

  24. [32]

    A normalized levenshtein dis- tance metric

    Li Yujian and Liu Bo. A normalized levenshtein dis- tance metric. IEEE transactions on pattern analysis and machine intelligence, 29(6) :1091–1095, 2007

    2007

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.