The ABC system is a domain-independent system for repairing faulty Datalog-like theories by combining three existing techniques: abduction, belief revision and conceptual change. Accordingly, it is named the ABC repair system (ABC). Given an observed assertion and a current theory, abduction adds axioms, or deletes preconditions, which explain that observation by making the corresponding assertion derivable from the expanded theory. Belief revision incorporates a new piece of information which conflicts with the input theory by deleting old axioms. Conceptual change uses the reformation algorithm for blocking unwanted proofs or unblocking wanted proofs. The former two techniques change an axiom as a whole, while reformation changes the language in which the theory is written. These three techniques are complementary. But they have not previously been combined into one system. We align these three techniques in ABC, which is capable of repairing logical theories with better result than each individual technique alone.
This repository contains two implementations of the ABC system, namely ABC_Datalog and ABC_FOL. ABC_Datalog is the original implementation of the ABC system, which is based on Datalog. ABC_FOL is the extension of ABC_Datalog, which is based on First-Order Logic (FOL).
Please refer to FOL_ABC_thesis.pdf for details such as the repair plans.
The folder evaluation contains the faulty theories tested in the evaluation in our project. The ones with name *h.pl is a theory with heuristics, while ones with name *nh.pl is the corresponding theory without heuristics.
example1.pl through example5.pl are the examples used in the thesis for ABC_FOL.
The theory file for the eggtimer example, as discussed in the CogAI2023 workshop, can be found in evaluation/eggtimer.pl and evaluation/eggtimer2.pl.
The implementations of ABC_FOL and ABC_Datalog are similar but independent. A corresponding folder in src contains the code for each implementation. There is a unified interface for calling the program of each implementation, which is main.pl. This is the file to be consulted when running the program.
Step1. Prepare the theory input file in another folder placed in the project root directory e.g., any file in folder evaluation. It has to include a Datalog theory given by axiom([...]), and the preferred structure given by trueSet([...]) and falseSet([...]). Then one can put the items to protect from being changed in protect([...]), and heuristics to apply in heuristics([...]). Add the following lines at the top of the theory input file:
:- working_directory(_, '../src').
:-[main].
logic(fol). %Choose between fol and datalog
theoryName(eggtimer). %Provide a name for your theory to identify the output files
Step2. In a prolog console, consult the theory input file, for example:
1 ?- working_directory(_,'./evaluation').
true.
2 ?- [mumh].
true.Step3. Call predicate abc. The output files include abc...faultFree.txt which contains the repair solutions; abc...record.txt which has the log information of ABC's procedure, and abc...repNum.txt which is the pruned sub-optimal.
- noAssAdd: no assersion will be added.
- noAxiomAdd: no axiom will be added, which equal to noAssAdd + no rules to be added.
- noAxiomDele: no axiom will be deleted.
- noPrecAdd: no precondition will be added to a rule.
- noPrecDele: no precondition will be deleted from a rule.
- noRuleChange: no repairs on rule changes.
- noAss2Rule: no assertion being repaired into a rule (An assertion will become a rule by being added a precondition)
- noCon2Rule: tbd
- noReform: no reformation repairs.
- noAnalogy: no rules will be produced based on the analogy algorithm.
- noRename: no rename a predicate/constant