Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair

Xue Li; Alan Bundy; Ricky Zhu

Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair

Xue Li^*, Alan Bundy, Ricky Zhu

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In general, axioms a logical theory are not equal in terms of their informational value (IV), and neither are the preconditions in a logical rule. Measuring IV is crucial, particularly in automated repair systems, e.g., the Abduction, Belief Revision and Conceptual Change (ABC) repair system \citep{li2022overview}, because when there are multiple repairs, the one that changes the item of least IV is preferred. However, quantifying IV is challenging. Given a benchmark, we
evaluate the IV from the perspective of how much an axiom/precondition supports the benchmark, which is quantitatively defined as the {\em vitality} of axioms and preconditions. The bigger contribution of an axiom/precondition in supporting the benchmark, the more information it conveys, so the bigger its vitality is. Our evaluation shows that the ABC repair system finds the best-repaired theories with smaller search space by only changing the least vital axioms/preconditions, as shown by our evaluation.

Original language	English
Title of host publication	Proceedings of the Tenth Annual Conference on Advances in Cognitive Systems
Editors	Jamie C. Macbeth, Leilani Gilpin, Michael T. Cox
Publisher	Computer Science: Faculty Publications
Pages	390-405
Number of pages	16
Publication status	Published - 13 Jan 2023
Event	The Tenth Annual Conference on Advances in Cognitive Systems - Arlington, United States Duration: 19 Nov 2022 → 22 Nov 2022 Conference number: 10 http://www.cogsys.org/conference/2022/

Conference

Conference	The Tenth Annual Conference on Advances in Cognitive Systems
Abbreviated title	ACS 2022
Country/Territory	United States
City	Arlington
Period	19/11/22 → 22/11/22
Internet address	http://www.cogsys.org/conference/2022/

Keywords / Materials (for Non-textual outputs)

Entrechment
Theory repair
Preferences

Access to Document

Benchmark based_LI_DOA_03112022_AFVAccepted author manuscript, 617 KBLicence: Creative Commons: Attribution (CC-BY)

https://scholarworks.smith.edu/csc_facpubs/366Licence: All Rights Reserved

1 Finished
1 Active

TEAMER : Teaching Machines to Reason Like Humans
Lapata, M. (Principal Investigator)
UK central government bodies/local authorities, health and hospital authorities
1/10/21 → 30/09/26
Project: Research
UKRI Trustworthy Autonomous Systems Node in Governance and Regulation
Ramamoorthy, R. (Principal Investigator), Belle, V. (Co-investigator), Bundy, A. (Co-investigator), Jackson, P. (Co-investigator), Lascarides, A. (Co-investigator) & Rajan, A. (Co-investigator)
EPSRC
1/11/20 → 30/04/24
Project: Research

Cite this

@inproceedings{ca9609b2f5144aeab79eb831f6f51071,

title = "Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair",

abstract = "In general, axioms a logical theory are not equal in terms of their informational value (IV), and neither are the preconditions in a logical rule. Measuring IV is crucial, particularly in automated repair systems, e.g., the Abduction, Belief Revision and Conceptual Change (ABC) repair system \citep{li2022overview}, because when there are multiple repairs, the one that changes the item of least IV is preferred. However, quantifying IV is challenging. Given a benchmark, we evaluate the IV from the perspective of how much an axiom/precondition supports the benchmark, which is quantitatively defined as the {\em vitality} of axioms and preconditions. The bigger contribution of an axiom/precondition in supporting the benchmark, the more information it conveys, so the bigger its vitality is. Our evaluation shows that the ABC repair system finds the best-repaired theories with smaller search space by only changing the least vital axioms/preconditions, as shown by our evaluation.",

keywords = "Entrechment, Theory repair, Preferences",

author = "Xue Li and Alan Bundy and Ricky Zhu",

year = "2023",

month = jan,

day = "13",

language = "English",

pages = "390--405",

editor = "Macbeth, {Jamie C.} and Leilani Gilpin and Cox, {Michael T.}",

booktitle = "Proceedings of the Tenth Annual Conference on Advances in Cognitive Systems",

publisher = "Computer Science: Faculty Publications",

note = "The Tenth Annual Conference on Advances in Cognitive Systems, ACS 2022 ; Conference date: 19-11-2022 Through 22-11-2022",

url = "http://www.cogsys.org/conference/2022/",

}

Li, X, Bundy, A & Zhu, R 2023, Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair. in JC Macbeth, L Gilpin & MT Cox (eds), Proceedings of the Tenth Annual Conference on Advances in Cognitive Systems. Computer Science: Faculty Publications, pp. 390-405, The Tenth Annual Conference on Advances in Cognitive Systems, Arlington, Virginia, United States, 19/11/22. <https://scholarworks.smith.edu/csc_facpubs/366>

Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair. / Li, Xue; Bundy, Alan; Zhu, Ricky.
Proceedings of the Tenth Annual Conference on Advances in Cognitive Systems. ed. / Jamie C. Macbeth; Leilani Gilpin; Michael T. Cox. Computer Science: Faculty Publications, 2023. p. 390-405.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair

AU - Li, Xue

AU - Bundy, Alan

AU - Zhu, Ricky

N1 - Conference code: 10

PY - 2023/1/13

Y1 - 2023/1/13

N2 - In general, axioms a logical theory are not equal in terms of their informational value (IV), and neither are the preconditions in a logical rule. Measuring IV is crucial, particularly in automated repair systems, e.g., the Abduction, Belief Revision and Conceptual Change (ABC) repair system \citep{li2022overview}, because when there are multiple repairs, the one that changes the item of least IV is preferred. However, quantifying IV is challenging. Given a benchmark, we evaluate the IV from the perspective of how much an axiom/precondition supports the benchmark, which is quantitatively defined as the {\em vitality} of axioms and preconditions. The bigger contribution of an axiom/precondition in supporting the benchmark, the more information it conveys, so the bigger its vitality is. Our evaluation shows that the ABC repair system finds the best-repaired theories with smaller search space by only changing the least vital axioms/preconditions, as shown by our evaluation.

AB - In general, axioms a logical theory are not equal in terms of their informational value (IV), and neither are the preconditions in a logical rule. Measuring IV is crucial, particularly in automated repair systems, e.g., the Abduction, Belief Revision and Conceptual Change (ABC) repair system \citep{li2022overview}, because when there are multiple repairs, the one that changes the item of least IV is preferred. However, quantifying IV is challenging. Given a benchmark, we evaluate the IV from the perspective of how much an axiom/precondition supports the benchmark, which is quantitatively defined as the {\em vitality} of axioms and preconditions. The bigger contribution of an axiom/precondition in supporting the benchmark, the more information it conveys, so the bigger its vitality is. Our evaluation shows that the ABC repair system finds the best-repaired theories with smaller search space by only changing the least vital axioms/preconditions, as shown by our evaluation.

KW - Entrechment

KW - Theory repair

KW - Preferences

M3 - Conference contribution

SP - 390

EP - 405

BT - Proceedings of the Tenth Annual Conference on Advances in Cognitive Systems

A2 - Macbeth, Jamie C.

A2 - Gilpin, Leilani

A2 - Cox, Michael T.

PB - Computer Science: Faculty Publications

T2 - The Tenth Annual Conference on Advances in Cognitive Systems

Y2 - 19 November 2022 through 22 November 2022

ER -

Benchmark based Vitality of Axioms and Preconditions for Datalog Theory Repair

Abstract

Conference

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Projects

TEAMER : Teaching Machines to Reason Like Humans

UKRI Trustworthy Autonomous Systems Node in Governance and Regulation

Cite this