pith. sign in

arxiv: 1906.09198 · v1 · pith:PIIERREEnew · submitted 2019-06-21 · 💻 cs.DB · cs.AI· cs.LO

Explainable Fact Checking with Probabilistic Answer Set Programming

Naser Ahmadi , Joohyung Lee , Paolo Papotti , Mohammed Saeed This is my paper

Pith reviewed 2026-05-25 18:20 UTC · model grok-4.3

classification 💻 cs.DB cs.AIcs.LO
keywords fact checkingknowledge graphsanswer set programmingprobabilistic inferenceexplainable AIlogical rule discoveryweb mining
0
0 comments X

The pith

A probabilistic answer set programming approach labels claims with explanations by combining knowledge graphs, rule discovery, and web mining.

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

The paper develops a fact checking method that assesses claims against knowledge graphs while generating explanations for its outputs. It compensates for incomplete graphs by discovering logical rules and mining web text to assemble evidence. Uncertain rules and facts are encoded as programs in a probabilistic extension of answer set programming. Inference over these programs produces both a true or false label for the claim and an explanation grounded in the logical structure. A reader would care because the method aims to make automated fact checking transparent rather than opaque.

Core claim

The paper claims that modeling the fact checking task as an inference problem in probabilistic answer set programming allows uncertain evidence from knowledge graphs, discovered logical rules, and web-mined facts to yield both accurate claim labels and human-interpretable explanations, with experimental results showing higher quality than state-of-the-art baselines.

What carries the argument

Probabilistic answer set programming, which encodes uncertain rules and facts from graphs and web sources into logical programs for inference that simultaneously labels claims and generates explanations.

If this is right

  • Claims receive both labels and explanations through probabilistic inference.
  • The method achieves higher quality results than existing baselines on the evaluated tasks.
  • Explanations draw directly on the semantic relationships stored in knowledge graphs.
  • Rule discovery and web mining together address gaps in the source graphs.

Where Pith is reading between the lines

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

  • The same encoding of uncertain evidence into answer set programs could support verification tasks in domains with structured but incomplete data beyond news claims.
  • Performance would likely degrade on claims requiring evidence types not well captured by logical rules or web text snippets.
  • Replacing the rule discovery step with learned embeddings might change both accuracy and the form of the generated explanations.

Load-bearing premise

Logical rule discovery combined with web text mining gathers sufficient evidence to assess claims despite the inevitable incompleteness of knowledge graphs.

What would settle it

A benchmark of claims where the system consistently fails to collect enough evidence from rules and web mining, resulting in either no label or explanations that contradict human judgments on claim validity.

Figures

Figures reproduced from arXiv: 1906.09198 by Joohyung Lee, Mohammed Saeed, Naser Ahmadi, Paolo Papotti.

Figure 1
Figure 1. Figure 1: Our Fact checking framework. p(x,y), we identify all the rules that have pred￾icate p and negp in the conclusion and the evi￾dence facts for the bodies of the rules. We then run LPMLN2ASP and check if p or negp are in the answer set. Problem Statement. Given an input claim to be verified and a KG, our goal is to compute an as￾sessment of the veracity of the claim and the expla￾nations for such decision, ex… view at source ↗
read the original abstract

One challenge in fact checking is the ability to improve the transparency of the decision. We present a fact checking method that uses reference information in knowledge graphs (KGs) to assess claims and explain its decisions. KGs contain a formal representation of knowledge with semantic descriptions of entities and their relationships. We exploit such rich semantics to produce interpretable explanations for the fact checking output. As information in a KG is inevitably incomplete, we rely on logical rule discovery and on Web text mining to gather the evidence to assess a given claim. Uncertain rules and facts are turned into logical programs and the checking task is modeled as an inference problem in a probabilistic extension of answer set programs. Experiments show that the probabilistic inference enables the efficient labeling of claims with interpretable explanations, and the quality of the results is higher than state of the art baselines.

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

0 major / 2 minor

Summary. The paper presents a fact-checking method that assesses claims using reference information from knowledge graphs (KGs) and generates interpretable explanations based on entity semantics and relationships. To handle inevitable KG incompleteness, it combines logical rule discovery with web text mining to gather evidence; uncertain rules and facts are encoded as programs, and claim verification is cast as inference in a probabilistic extension of answer set programming. Experiments are reported to show that the approach enables efficient claim labeling together with explanations and yields higher quality than state-of-the-art baselines.

Significance. If the experimental claims hold, the combination of probabilistic ASP inference with explicit rule discovery and web mining supplies a concrete route to transparent fact checking that respects both logical structure and uncertainty; the explicit acknowledgment of KG incompleteness and the positioning of rule discovery as mitigation are strengths that could influence subsequent work on explainable verification systems.

minor comments (2)
  1. [Abstract] Abstract: the phrase 'probabilistic extension of answer set programs' is used without naming the concrete formalism (e.g., ProbLog, P-log, or a custom semantics); a one-sentence clarification would help readers locate the exact inference engine.
  2. [Experiments] The manuscript should include a short table or paragraph that lists the concrete datasets, number of claims, and baseline systems used in the reported experiments so that the 'higher quality' claim can be directly compared with prior work.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our paper and the recommendation for minor revision. The review correctly highlights the integration of probabilistic ASP inference with rule discovery and web mining as a strength for transparent fact checking that accounts for KG incompleteness.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper frames fact checking as inference in probabilistic ASP after explicit steps of KG lookup, rule discovery, and web mining to address incompleteness; the abstract states these choices directly without any derivation that reduces a claimed prediction or uniqueness result to a fitted parameter or prior self-citation by construction. No equations, ansatzes, or load-bearing citations are exhibited that collapse the central result to its inputs. The experimental comparison to baselines remains an independent evaluation step.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the domain assumption that knowledge graphs provide usable semantic structure and that rule discovery plus web mining can compensate for incompleteness; no free parameters or invented entities are described in the abstract.

axioms (2)
  • domain assumption Knowledge graphs contain a formal representation of knowledge with semantic descriptions of entities and their relationships.
    Stated directly in the abstract as the foundation for using KGs.
  • domain assumption Information in a KG is inevitably incomplete, requiring external evidence from rule discovery and web mining.
    Explicit premise used to justify the need for probabilistic handling.

pith-pipeline@v0.9.0 · 5676 in / 1044 out tokens · 22709 ms · 2026-05-25T18:20:11.477397+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

  1. [1]

    Babakar and W

    M. Babakar and W. Moy. The state of automated factchecking. https: //fullfact.org/blog/2016/aug/ automated-factchecking/, 2016

    work page 2016

  2. [2]

    Bordes, N

    A. Bordes, N. Usunier, A. Garcia-Duran, J. Weston, and O. Yakhnenko. Translating embeddings for modeling multi-relational data. In NIPS, pages 2787–2795, 2013

    work page 2013

  3. [3]

    T. D. Cao, I. Manolescu, and X. Tannier. Searching for truth in a database of statistics. In WebDB, pages 4:1–4:6, 2018

    work page 2018

  4. [4]

    G. L. Ciampaglia, P. Shiralkar, L. M. Rocha, J. Bollen, F. Menczer, and A. Flammini. Computational fact checking from knowl- edge networks. PloS one, 10(6), 2015

    work page 2015

  5. [5]

    X. Dong, E. Gabrilovich, G. Heitz, W. Horn, N. Lao, K. Murphy, T. Strohmann, S. Sun, and W. Zhang. Knowledge vault: a web-scale approach to probabilistic knowledge fusion. In KDD, pages 601–610, 2014

    work page 2014

  6. [6]

    M. H. Gad-Elrab, D. Stepanova, J. Urbani, and G. Weikum. Exfakt: A framework for explaining facts over knowledge graphs and text. In WSDM, pages 87–95, 2019

    work page 2019

  7. [7]

    Gal ´arraga, C

    L. Gal ´arraga, C. Teflioudi, K. Hose, and F. M. Suchanek. Fast rule mining in ontological knowledge bases with AMIE+. The VLDB Journal, 24(6):707–730, 2015

    work page 2015

  8. [8]

    Gatt and E

    A. Gatt and E. Krahmer. Survey of the state of the art in natural language generation: Core tasks, applications and evaluation.Jour- nal of Artificial Intelligence Research, 61:65– 170, 2018. 10

    work page 2018

  9. [9]

    Hassan, F

    N. Hassan, F. Arslan, C. Li, and M. Tremayne. Toward automated fact- checking: Detecting check-worthy factual claims by claimbuster. In KDD, 2017

    work page 2017

  10. [10]

    Huynh and P

    V . Huynh and P. Papotti. Towards a benchmark for fact checking with knowl- edge bases. In Companion of TheWebConf (WWW), pages 1595–1598, 2018

    work page 2018

  11. [11]

    Jaradat, P

    I. Jaradat, P. Gencheva, A. Barr ´on-Cede˜no, L. M`arquez, and P. Nakov. ClaimRank: De- tecting check-worthy claims in Arabic and English. In NAACL-HTL, pages 26–30, 2018

    work page 2018

  12. [12]

    J. Leblay. A declarative approach to data- driven fact checking. In AAAI, pages 147– 153, 2017

    work page 2017

  13. [13]

    J. Lee, S. Talsania, and Y . Wang. Comput- ing LPMLN using ASP and MLN solvers. Theory and Practice of Logic Programming , 2017

    work page 2017

  14. [14]

    Lee and Y

    J. Lee and Y . Wang. Weighted rules under the stable model semantics. In KR, pages 145– 154, 2016

    work page 2016

  15. [15]

    Lehmann, D

    J. Lehmann, D. Gerber, M. Morsey, and A. N. Ngomo. Defacto - deep fact validation. In ISWC, pages 312–327, 2012

    work page 2012

  16. [16]

    Ortona, V

    S. Ortona, V . V . Meduri, and P. Papotti. Ro- bust discovery of positive and negative rules in knowledge bases. In ICDE, pages 1168– 1179, 2018

    work page 2018

  17. [17]

    Popat, S

    K. Popat, S. Mukherjee, J. Str ¨otgen, and G. Weikum. Credibility assessment of textual claims on the web. In CIKM, 2016

    work page 2016

  18. [18]

    Popat, S

    K. Popat, S. Mukherjee, A. Yates, and G. Weikum. Declare: Debunking fake news and false claims using evidence-aware deep learning. In EMNLP, pages 22–32, 2018

    work page 2018

  19. [19]

    Richardson and P

    M. Richardson and P. Domingos. Markov logic networks. Machine learning , 62(1- 2):107–136, 2006

    work page 2006

  20. [20]

    Shi and T

    B. Shi and T. Weninger. Discriminative predicate path mining for fact checking in knowledge graphs. Knowledge-Based Sys- tems, 104:123–133, 2016

    work page 2016

  21. [21]

    Shiralkar, A

    P. Shiralkar, A. Flammini, F. Menczer, and G. L. Ciampaglia. Finding streams in knowl- edge graphs to support fact checking. In ICDM, pages 859–864, 2017

    work page 2017

  22. [22]

    Socher, D

    R. Socher, D. Chen, C. D. Manning, and A. Ng. Reasoning with neural tensor net- works for knowledge base completion. In NIPS, pages 926–934, 2013

    work page 2013

  23. [23]

    Srinivasan

    A. Srinivasan. The Aleph manual, 2001

    work page 2001

  24. [24]

    F. M. Suchanek, G. Kasneci, and G. Weikum. Yago: a core of semantic knowledge. In WWW, pages 697–706. ACM, 2007

    work page 2007

  25. [25]

    Thorne, A

    J. Thorne, A. Vlachos, C. Christodoulopou- los, and A. Mittal. FEVER: a large-scale dataset for fact extraction and verification. In NAACL-HLT, 2018

    work page 2018

  26. [26]

    Fairness and Explainability: From ideation to implementation

    Workshop. Fairness and Explainability: From ideation to implementation. In NeurIPS, 2019

    work page 2019

  27. [27]

    Y . Wu, P. K. Agarwal, C. Li, J. Yang, and C. Yu. Toward computational fact- checking. Proceedings of the VLDB Endow- ment, 7(7):589–600, 2014. 11

    work page 2014